European Historical Economics Society
Transcript of European Historical Economics Society
European Historical Economics Society
EHES Working Paper | No. 153 | May 2019
Trade in the Shadow of Power: Japanese Industrial Exports in the Interwar years
Alejandro Ayuso-Díaz
Antonio Tena-Junguito
Universidad Carlos III
1 Departamento de Ciencias Sociales, Universidad Carlos III, Calle Madrid, 126, 28903 Getafe, Spain. E- mail: [email protected] 2 Departamento de Ciencias Sociales, and Researcher at Instituto Figuerola, Universidad Carlos III, Calle Madrid, 126, 28903 Getafe, Spain. E- mail: [email protected] http://portal.uc3m.es/portal/page/portal/dpto_ciencias_sociales/profesorado/antonio_tena
EHES Working Paper | No. 153 | May 2019
Trade in the Shadow of Power: Japanese Industrial Exports in the Interwar years
Alejandro Ayuso-Díaz1
Antonio Tena-Junguito2
Universidad Carlos III
Abstract
During the interwar years, Japanese industrialisation accelerated alongside the expansion of industrial exports to regional markets. Trade blocs in the interwar years were used as an instrument of imperial power to foster exports and as a substitute for productivity to encourage industrial production. The historiography on Japanese industrialisation in the interwar years describes heavy industries’ interests in obtaining access to wider markets to increase economies of scale and reduce unit costs. However, this literature provides no quantitative evidence that proves the success of those mechanisms in expanding exports. In this paper we scrutinise how Japan—a relatively poor country—used colonial as well as informal power interventions to expand regional markets for its exports, especially for the most intensive human capital sector of the industrializing economy. JEL Codes: F14, N15, N75 Keywords: International Trade, Empires, Trade blocs, Japan, Interwar years
Notice
The material presented in the EHES Working Paper Series is property of the author(s) and should be quoted as such. The views expressed in this Paper are those of the author(s) and do not necessarily represent the views of the EHES or
its members
1
Introduction
This paper studies the main drivers of the expansion of Japanese exports during
the interwar years at a time of accelerated growth and convergence. Japan underwent a
process of rapid industrialization on the basis of import substitution and imperial policies.
Its newly created empire was very efficient in developing a peculiar imperial trade in the
shadow of power throughout East and Southeast Asia in conjunction with a more
aggressive imperial regional policy through conquest.1
Of course, such protectionist and imperial policies in the international economy
were not confined to Japan during the interwar years. Britain, for example, instituted a
policy of imperial preference with its colonies and mandated territories in the framework
of the Commonwealth at the Ottawa Conference of 1932. Germany, having neither
complete customs autonomy as a result of the Treaty of Versailles nor an informal or
formal geographically contiguous empire, fostered a complex and domineering system of
clearing arrangements with Central European and Balkan neighbours even before the
Nazi regime. Trade barriers rose all over the world and the biggest economies abandoned
1 Trade has taken place in the shadow of power for all almost all of recorded human history. We argue that
power matters for trade, as critically as the traditional determinants of endowments, preferences, and
technology Following Findlay and O’Rourke, 2007 and Garfinkel. et al (2012).
2
global multilateral free trade policies in favour of commercial exchanges within their
Empires. Fragmented regional or global empires, both formal and informal, were
reconciled ambivalently with bilateral agreements that included tariffs, quotas, import
licenses, exchange controls, barter and clearing agreements, and other protectionist
policies.
Japanese imperialist history is well known. During the Meiji Period (1868-1912),
Japan annexed Ezo (Hokkaidô) in 1869, the Ryûkyû Islands in 1879, Taiwan in 1895, and
Korea between 1905-10. Until the 1920s, Japan allowed free access to foreigners in their
occupied territories and only then created a real imperial bloc in conjunction with an
escalation of domestic trade barriers. By the early 1930s, Japan already had consolidated
a decisive policy of protectionism and an empire wide regional market with trading
privileges for Japanese industrial firms. Tariff policy in the Japanese occupied territories
privileged Japan or was assimilated into Japan’s own tariff system. Korea adopted Japan’s
tariff system in 1923, as would Manchuria ten years later upon the territory’s separation
from the Chinese custom system, allowing preferential access for Japanese
manufactures.2
Along with the protectionist backlash experienced around the world, the
development of the Japanese Imperial Bloc in the interwar years also responded to the
demand of Zaibatsu heavy industrialists for exclusive access to wider markets and the
military’s desire to control vital strategic materials, such as oil, rubber, and iron ore. These
interests were consolidated by the conquest of Manchuria as well as the regional
expansion of commercial and investment networks in other areas that were later called
‘The Greater East Asia Co-Prosperity Sphere’.3
This paper will assess how much of the expansion of Japan’s industrial exports
during the interwar years was based on this combination of imperial soft power and its
policy of aggressive conquest in East and Southeast Asia. Specifically, we examined
whether the expansion of high-skill exports to Japan’s neighbours can be better explained
by improvements in productivity or by imperial policy variables. For that purpose, we
ascertained the mechanisms that linked Japanese imperial activity to its exports. We
defined the ‘Japanese Empire’ (JE) as including both previously occupied colonies and
2 See Chase (2005, 62-64).
3 See Fletcher (1989, 144-50). Similar to the term ‘Third Reich’, this was one of a number of slogans and
concepts used in the justification of Japanese aggression in East Asia from the 1930s through the end of
World War II.
3
what we call ‘Future Conquests’—those countries or polities in the region that were of
strategic interest for Japan, as proven by their occupation between 1941 and 1945.4 We
argue that Japan exerted regional commercial influence through the creation of
transnational commercial and investment networks and cultural diplomacy, which
facilitated commercial information and created merchant community trust in the region.
This soft power thrust, combined with a more traditional imperial policy of military
conquest, was decisive for the consolidation of a new stage of industrial modernization
during the 1930s.
To study the determinants of Japanese exports, we digitized Japanese commercial
returns (Annual Returns of the Foreign Trade of the Empire of Japan) for several
benchmark years (1912, 1915, 1925, 1929, 1932, and 1938) and created an exhaustive
granular product data set of exports by country of destination for each of the chosen years.
Using this data set we analysed the main determinants of Japanese exports by skill bias,
focusing mainly on East and Southeast Asia, and comparing them with Japan’s export
performance with the rest of the world. The methodology employed for that purpose is in
line with extensive and intensive marginal gravity models. The specification used in this
paper includes the usual factor endowments, demand structure and trade cost variables
(such as tariffs, transport cost, and political factors). The basic gravity model is
augmented by a new series of freight factors from Japan to corresponding destinations
(disaggregated by product line), commercial diplomacy appointments, and dependency
relationships with Japan (occupation, annexation, colonisation, and lease) before and after
1932.
Japan’s second stage of industrial expansion in the interwar years was mainly
based on exports of high-skill manufactured goods, and this expansion was fostered by
the imperial shadow of power. The literature on Japanese industrialisation in the interwar
years describes heavy industries’ interests in gaining access to wider markets to increase
economies of scale and reduce unit costs. Yet, there exists no quantitative evidence that
proves the success of those mechanisms in expanding exports to regional markets. In this
paper we use a new data set and other empirical evidence to answer the following
4 Formal colonies include: Taiwan, Korea, Kwantung Leased Territory, Manchuria (1932), and China
(1938). The “Future Conquests” are: Manchuria (1912-1929), China (1912-1932),Thailand, Burma, French
Indochina, Hong Kong, Dutch East Indies, British Borneo and Sarawak, New Guinea, British Malaya
(including Singapore), the Philippine Islands, and the Solomon, Gilbert, and Marshall Islands in the Pacific.
For more information on the main features of the colonies and future conquest territories, see Section 4.3,
Appendix A.
4
questions. First, were improvements in Japanese relative productivity or comparative
advantage the main drivers of its industrial export expansion? Or, conversely, was trade
expansion driven mainly by political enforcement related to imperial privileges? Our
hypothesis, in short, is that the Imperial Bloc was a much stronger driver of Japanese
export expansion than the conventional reduction of trade costs or improvements in
comparative advantage. Japan reacted to the interwar context of global commercial
disintegration with aggressive import substitution and an imperial strategy to foster
exports of manufactures, as did Britain and Germany in other ways.
2. Regional Trade and the Japanese Empire
What mechanisms link imperial expansion with increases in exports towards a
determined region? Either international context or domestic industrial policy helps to
understand the complex Japanese interest in East and Southeast Asia, and its expansionary
policy of trade in the shadow of power.5
The British commercial relationship with its colonies after 1932, especially with
India, with Britain turning away from almost a century of free trade policy, is probably
the most well-known case study of the use of trade in the shadow of power. However,
most of the literature presents the Commonwealth trade bloc as being less discriminatory
towards outsiders than initially suggested by the text of the Ottawa agreement of July
1932. Nevertheless, the agreement succeeded in implementing preferential access by
Britain to the dominions and vice versa. Great Britain imposed a general tariff of 10% for
manufactures and other goods, and increased the Empire’s share of British imports by
approximately 70% between 1930 and 1933 (De Bromhead et al. (2017). The British
partially closed down Indian, African, and other imperial markets, into which some
competitive Japanese textiles had expanded since the end of the 19th Century. Certainly
the Manchurian invasion of 1931 preceded Ottawa, but this international context
influenced Japanese expansionist commercial policy towards its empire during the 1930s
nevertheless (See discussion in Appendix D).6
The Third Reich followed other practices emulating policies of trade in the
shadow of power during this period, and Germany consolidated its imperial ambitions in
5Garfinkel, M. et al (2012). ‘Trade in the Shadow of Power’; Eichengreen, B., & Irwin, D. A. (1995). 6 See also Cain, P. J., & Hopkins, A. G. (1980); Gallagher, J., & Robinson, R. (1953).
5
the Balkans and Central Europe during the Nazi regime. Germany used political
enforcement and commercial influence on their neighbours to guarantee those primary
and mineral resources necessary for geostrategic military considerations (Milward (1981)
and Ritschl (2001)). During the 1930s, German trade was facilitated by bilateral clearing
agreements, mainly signed with countries in Central and Southeastern Europe, but also in
South America. Under this new regime, German purchases were credited against
purchases by foreigners in German markets.7
The case of Japanese imperial policy is peculiar not only because it predated
World War I but also, most importantly, because it was developed by a relatively poor
developing country and was restricted to a regional framework.8 In addition to its formal
colonies, Japanese economic agents built intense business and political networks with
other countries in East and Southeast Asia. The motive behind this activity was largely
commercial and was manifested in the subsequent military occupations, during the
Second World War, of Thailand, Burma, French Indochina, Hong Kong, the Dutch East
Indies, British Borneo and Sarawak, New Guinea, British Malaya (including Singapore),
the Philippine Islands, and the Solomon, Gilbert, and Marshall Islands in the Pacific.
The literature on Japanese colonialism has mainly focused on colonial trade links
with Japan after occupation, although Kublin (1959) and Duus (1998) also observed that
the Japanese demonstrated their interest in Korean markets when they forced the opening
of Korean ports and signed the unequal commercial treaties of 1876. After occupation in
1905, colonial trade policy was directed by Tokyo in order to ensure the safe supply of
foodstuffs and raw materials to Japan and to guarantee a market for Japanese
manufactures. This colonial policy was modified after the global rise of protectionism
following the Great Depression. Japan pursued growth with an empire-wide economic
strategy. The annexation of Manchuria in 1931 and the adoption of an import substitution
strategy signalled the beginning of Japanese planned industrialisation within its colonies.9
This strategy generated an important increase in the production of high-skilled
manufactures, like steel for rail construction and machinery, that was directed mainly
towards the Japanese domestic market, but was also exported to foster Korean
7Neal, L. (1979); Gross, S. (2016).
8 Matsusaka, Y. T. (2007).
9 Kohli, A. (1994).
6
industrialisation.10
Another mechanism granting Japanese exports access to its colonies was the fact
that the Taiwanese and Korean economies were integrated with Japan by the removal of
trade barriers and the introduction of fixed exchange rates.11 Additionally, Japan invested
in the transport and communication infrastructure and imposed the assimilation of the
Japanese language, which further reduced transaction costs. Commercial exchange was
further facilitated by private sector investment and the establishment of Japanese
companies (in fact, almost every company in Korea was owned by Japanese) along with
the opening of Japanese financial institutions like the Bank of Chosen.12
The above scenario is what happened in the best known Japanese colonies.
However, the Japanese Empire also included two other territories, which provide relevant
examples. The most relevant case was Manchuria, which became a Japanese puppet state
in 1931, and was considered an additional Japanese colony. The Japanese had a presence
in Manchuria prior to annexation, however, and this presence became more pronounced
during the First World War. Japanese economic activity consisted mainly in the
establishment of large companies (such as banks, similar to the Bank of Chosen), of
raising FDI flows, and of increasing Japanese private investment in transport
infrastructure (aided by Government subsidies). These activities permitted the Japanese
to dodge most trade barriers and to reduce other transaction costs.13 Eckstein et al. (1974)
suggested that Japanese investment before 1930 fostered Manchurian industrialisation.
After the Japanese annexation of Manchuria, a puppet state was created with the primary
objective of influencing the direction and materials necessary for the rapid
industrialisation of the region.
The second relevant case is that of the Kwantung Leased Territory, commonly
presented as the most representative example of the ‘conquer by railway’ strategy. In
1906, after acquiring the territory during the Russo-Japanese War, the Japanese
government founded the South Manchurian Railway company (SMR) to operate the
railway network left by the Russians. By 1930 it had become the largest joint stock
10 Chenery, H. B., Shishido, S., & Watanabe, T. (1962).
11 The Taiwanese and Korean currencies were pegged at the same value to the yen from 1895 to 1946
(Taiwan) and from 1910 to 1945 (Korea) , see Schuler, K. (2004). 12 Aziz, M. A. (2012).
13Bix, H. (1972); Encarnation, D. J. (ed.) (1999).
7
company operating in the Japanese Empire. The company diversified into other activities
apart from railway construction, but it remained a strong source of demand for Japanese
goods like iron, machinery, and transport equipment used for mining, manufacturing, and
railway construction.14
The example of Manchuria illustrates the singular nature of the Japanese imperial
shadow of power: a process in which Japan established economic links with neighbouring
territories as a demonstration of interest or announcement of future military occupation.
We follow this historical case to justify the inclusion of a sample of countries and polities
as part of Japan’s informal empire before they were occupied during WWII (‘future
conquest’). It is worth distinguishing between those territories previously colonised by
European powers and those that were not. In the former case, Japanese investment was
severely restricted by Europeans, whereas in the latter, Japanese investment
diversification permitted an expansion of commercial relations.
Manchuria was the country in which Japan had stronger economic penetration..
Japan’s presence was characterised by the establishment of Japanese Zaibatsu (main
industrial conglomerates) and banks that served to elude trade barriers and to reduce
transaction costs.15 In fact, the operation of Japanese economic interests was reflected not
only in Japanese exports to China but also was related to Japanese military campaigns in
its neighbouring country (which prefigured the origins of Japanese conflict with the USA
and Great Britain).16
Furthermore, Japanese economic penetration took a different shape in those places
where Japan could not implement huge investments, like British Malaya, the Dutch East
Indies, or the Philippines. In these areas, Japanese activity was mainly characterised by
aggressive marketing strategies that were possible due to precise knowledge of the
cultural and geographic characteristics of these regional markets.17 Moreover, Japan
fostered efficient distribution channels for allocating Japanese products, taking advantage
14Iyenaga, T. (1912).
15 Howe, C. (1999); Osterhammel, J. (1986).
16 He (2007) mentions the complicity between Zaibatsu and the military power. Howe (1999), however,
denies full agreement between both parties by pointing out that although Japanese businessmen asked for
diplomatic sanctions against China in response to the Chinese boycott of Japanese goods, the Japanese
military aim of transforming China into a population outlet during the 1930s was at odds with the Japanese
business community in China, which only looked for political stability and business opportunities.
17 Post, P. & Lindblad, J. T. (1996).
8
of good commercial relationships with local merchants. A final characteristic of Japan’s
economic links with Southeast Asia took the form of massive Japanese immigration to
territories like the Philippines, with Japanese expatriates working in the trading sector,
with the majority of these expatriates demanding Japanese goods.18
The historiography suggests that Japanese geostrategic and economic interests
initially contemplated a form of economic penetration by exports and investments in
territories where its economic interests would be later unequivocally confirmed by
annexation during World War II. Our paper contributes to the literature on trade in the
shadow of power by highlighting Japan’s singularities and by providing an empirical
demonstration of the success of the mechanisms linking the Japanese Empire with
regional exports in East and Southeast Asia.
3. Empire and the Determinants of Japanese Exports
The late-19th and early-20th centuries represented a continuation of the period of
industrialisation in Japan, which began in the 1850s after the forced opening of the ports
by Western powers. This process permitted an important expansion of Japanese GDP per
capita, as is observed in Figure 1. The expansion of exports was an important component
of this remarkable economic growth, as manufacturing exports were one of the main
drivers of such good comparative performance.19 The dynamic behaviour of
manufacturing exports contrasts with that of primary products, which remained almost
stagnant in per capita terms for 60 years. From the end of the nineteenth century, textiles
and other manufactures were the main commodities exported (see Howe (1999) and
Sugiyama (2013)).20 This process accelerated in the 1930s, with an expansion of
manufacturing exports outstanding in relation to world export. This expansion of exports
consolidated industrial modernization and the historical change of comparative advantage
away from primary exports.
18 Shiraishi, T. (1993); Fisher, C. (1950); and Koh Soo Jin, D., & Tanaka, K. (1984).
19 See Meissner & Tang (2017).
20
Taking into account that raw silk is sold in bulk, whereas silk manufactures are sold individually, and
that most transformations needed to produce raw silk are hand-made, we include raw silk as a primary
product in Figure 1, following Sugiyama (2013), Francks (2015), Smitka (1998), and Souza (2004).
9
Figure 1: Japanese Exports (1934-36 $) and GDP per Capita (1913 $) and Japanese
and World Manufacturing Export Performance (1953=100)
Source: Constant exports come from Ohkawa et al. (1967-1989) Long Term Economic Statistics of Japan
(LTES) and GDP per Capita in 1990 Int. dollars from Bolt, J. and van Zanden, J. L. (2014). Export shares
come from Annual Returns of the Foreign Trade of the Empire of Japan and Japan and World comparative
manufacture exports in volume (1953=100) from UN Historical Trade Statistics.
Japanese industrialisation apparently generated increases in productivity that were
behind the observed rise in manufacturing exports. As mentioned in the previous section,
however, during the interwar years, Japan expanded imperial networks with its colonies
in order to develop a market for domestic industrial exports. In that sense, we will offer
some preliminary evidence to clarify whether increases in productivity or imperial
mechanisms exerted a stronger influence on the expansion of manufacturing exports.
3.1 Productivity or Empire, Which was More Relevant?
The first thing to be highlighted is that, as expected, Japanese overall productivity
increased importantly after WWI. Figure 2 shows that this hypothesis holds true
regardless of whether one proxies productivity by tertiary education, real wages, or
manufacturing output per hour worked. Nevertheless, the data also shows that Japanese
productivity stagnated during the period 1932-38, which is the period in which exports
experienced their greatest boom.
0
500
1000
1500
2000
2500
3000
0
5
10
15
20
1870 1880 1890 1900 1910 1920 1930
TOTAL EXPORTS PER CAPITA 1934-36 DOLLARS (LEFT AXIS)
MANUFACTURING EXPORTS PER CAPITA (1934-36 DOLLARS)
PRIMARY EXPORTS PER CAPITA 1934-36 DOLLARS
GDP per Capita (Right axis)
0
50
100
150
200
250
300
1900 1904 1908 1912 1923 1927 1931 1935
World Export Manufac Japan Export Manufact
10
Figure 2: The Evolution of Japanese Productivity and Exports (1912-1938)
Note: Exports refer to current exports in yen divided by the corresponding exports price index (base 1913).
The higher education ratio is the percent of tertiary education students over total population. Industrial
real wages refer to wages in manufacturing. Manufacturing output per hour refers to output in constant
1929 dollars divided by hours worked in manufacturing.
Source: Exports and Manufacturing output per hour come from data set, see Section 4. Education levels
are from Mitchell (2003). Finally industrial wages come from Ohkawa et al. (1967-1989).
The 1930s stagnation of manufacturing productivity in comparative terms also is
highlighted by Broadberry et al. (2015), who showed a process of Japan’s catching up
with the UK in terms of manufacturing productivity between the 1890s and 1920s and
that this process relented during the 1930s and 1940s. This means that increases in
productivity would better explain the expansion of exports before 1929 than after, when
other factors would have a higher incidence of helping to understand export
performance.21
On the other hand, Figure 3 shows that during the period of the faster growth of
Japanese manufacturing exports, the main destinations of these exports were within the
Japanese sphere of influence. More than 50% of exports after WWI went to territories
within this sphere. (Here we include both colonies that received the majority of
21 There is a remote possibility that past increases in productivity are related with future increments of
exports. Nevertheless, our regressions deny this possibility by including a lag of productivity as a likely
determinant of current exports, and past increases in productivity were not significant in any specification.
0
50
100
150
200
250
300
0
200
400
600
800
1.000
1.200
1.400
1.600
1.800
2.000
1912 1915 1925 1929 1932 1938
Exports constant (lett axis/ millons $ 1913)
Higher education Index (1912=100)
Indutrial real wages Index(1912=100)
Japan Manufacturing Output per hour index (1912=100)
11
manufacturing exports and territories in Southeast Asia occupied during WWII.) Further,
these territories’ share of Japanese exports rose faster than for destinations outside of
Japan’s sphere of influence during the 1930s, reaching 73% of the total in 1938.
Figure 3: Japanese Manufacturing Exports by Region (1912-1938) (%)
Note: The remaining share of 100% corresponds to poor countries outside Asia.
Source: Annual Return of Trade for the Empire of Japan (various years).
Certainly, the international context in the interwar years had changed drastically,
especially after the Great Depression, and domestic and regional market expansion was
conditioned by Europe’s and the United States’ significant increase in manufacturing
trade barriers and the development of trade blocs. What happened in the Indian market,
where Japan had increased its share of textile exports in competition with Britain, is a
clear example of the trade diversion forced by these policies. Japan was forced to retreat
from those territories after the Commonwealth’s trade bloc policy, adopted after the 1932
Ottawa conference, raised tariffs (for an extended discussion of this point, see Appendix
D).
Japan’s remarkable success in manufacturing exports during 1932-38 is not
comparable to any other country’s experience during the period (see Figure 1). Evidence
of improvements in productivity does not explain such success. We think this success is
principally determined by the reinforced bias of manufactures towards countries under
Japanese imperial influence. This evidence may indicate that the effect of colonial
mechanisms and other regional commercial influences with the future occupied territories
previously mentioned were more relevant than productivity increases. The next section
analyses how those mechanisms worked and what kind of exports they facilitated.
63%47%
65% 63% 56%73%
10%28%
17% 18% 24%11%
26% 24% 14% 13% 11% 8%
0%10%20%30%40%50%60%70%80%90%
100%
1912 1915 1925 1929 1932 1938
% Rich Countries %Other Asia and Pacific % Imperial
12
3.2 Did Empire Facilitate New Exports or Reinforce Pre-Existing Comparative
Advantage?
The first thing we must disentangle is the pattern of Japanese comparative
advantage. For that purpose, Table 1, based on Howe (1999), shows that among industrial
products, Japanese exporters enjoyed a strong comparative advantage in textiles and
clothing, which were low-skilled in nature.
Table 1: Revealed Comparative Advantage by Country in Three Main Commodities
TEXTILES/CLOTHING 1899 1913 1929 1937 BRITAIN 1.27 1.42 1.48 1.56 U.S.A 0.21 0.23 0.27 0.21 FRANCE 1.05 1.25 1.38 1.14 GERMANY 0.74 0.55 0.58 0.49
JAPAN
1.67 1.79 2.45
2.7
MACHINERY/ELECTRICAL EQUIPMENT BRITAIN 1.19 0.93 0.8 0.93 U.S.A 2.32 1.85 1.72 1.65 FRANCE 0.39 0.34 0.46 0.43 GERMANY 0.86 1.28 1.31 1.29 JAPAN 0 0.04 0.15 0.34 MACHINERY/ TRANSPORT EQUIPMENT BRITAIN 1.68 1.19 0.92 0.96 U.S.A 1.17 1.37 2.07 2.2 FRANCE 0.47 1.25 0.68 0.72 GERMANY 0.49 0.7 0.4 0.75 JAPAN 0 0.13 0.13 0.39 Note: Revealed comparative advantage is defined as a country's share in world exports of a particular manufactured commodity, divided by its share in the respective world manufacturing exports. Source: Howe (1999).
Japan’s comparative advantage in textiles was a relevant part of the Japanese drive
to increase manufacturing exports before the Second World War. Textiles are included in
our low-skill sector (which we divide between high-end and low-end, see Appendix C)).
Most of the textiles exported were low-end, representing a stable share around 70%
throughout the interwar years. Before the First World War, textiles were mainly exported
to formal and informal Japanese regions of influence—mostly from Hong Kong, China,
13
and Korea (see Meissner and Tang (2018)). But they spread to other destinations outside
the Japanese sphere of influence during the 1920s, such as India. A change in the
destinations of textile exports occurred after the Ottawa agreement of 1932, when textile
exports directed to India, were again reallocated to Japanese colonies (more evidence on
this point is developed in Appendix D).
This pattern is also reflected in the composition of Japanese exports during the
interwar years, which was mainly focused on low-skilled manufactures, as shown in
Figure 4. The figure also highlights that high-skilled exports, which did not enjoy
comparative advantage, led export expansion during the period 1932-38. This fact
suggests that such expansion was possible largely because these goods were directed to
countries inside the Japanese sphere of influence.
Figure 4: Japanese Total Manufacturing Exports by Skills and Region.
a) Manufacturing Exports by Skills b) Real High-Skill Exports by
Region
a) low and high skill b) high skilled exports by reg
Source: See data set, Section 4, and Annual Returns of the Foreign Trade of the Empire of Japan.
Next, it is pertinent to provide specific examples of how imperial mechanisms
fostered Japanese high-skilled exports within East and Southeast Asia.22 Regarding
Japanese colonies, it has been previously mentioned that colonial economies were
22Our data shows that Japanese colonies received a remarkable proportion of infrastructure and military
related high-skilled exports (iron and steel for railways, cement for railroads, weapons and munitions).
Those goods were mainly demanded by Japanese military colonisers. We also see, however, the rising
relevance of machinery and other industrial exports devoted to the economic development of the colonies,
which was demanded by Japanese businessmen established there. Some authors like Kohli, A. (1994) relate
those exports to Korean future development.
63% 58%72% 67% 66%
56%
25%27%
25% 32% 31%42%
11% 15%3% 1% 3% 2%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1912 1915 1925 1929 1932 1938
Low Skilled High Skilled Medium Skilled
43%46%
56% 63%65%
86%57% 54%
44% 37% 35%14%
0
200
400
600
800
1000
0%
20%
40%
60%
80%
100%
1912 1915 1925 1929 1932 1938
COLONIES FUTURE CONQUESTS
REST OF WORLD REAL HS EXPORTS
14
strategically directed from Tokyo, and that a rapid industrialisation plan was launched in
Korea and Manchuria during the 1930s. Figure 5 shows how this plan facilitated the
expansion of Japanese high-skilled exports. After 1925, Japan reduced its machinery
imports and began a massive increase of local machinery production, which was
completed by 1938. At the same time, the percentage of exported machines over total
production exploded after 1925, suggesting that this import substitution strategy was
complemented by the necessary extension of exportation of machinery to other territories
in order to consolidate the Japanese industrialisation process. In that sense, the right hand
portion of Figure 5 shows that those new markets were located in Japanese colonies,
which received 80% of total machinery exports in 1938. So, apparently, the
industrialisation of the colonies was based on Japanese machinery exports, opening new
markets for Japanese producers.
Figure 5: Index of Machinery Exports, Imports, and Domestic Production
(1912=100)
Note: Indexes measure the evolution of the shares of exported, imported, and consumed machinery over
total Japanese production from 1912.
Source: Ohkawa et al. (1967-1989) and Annual Returns of the Foreign Trade of the Empire of Japan.
Furthermore, it is also crucial to understand how economic connections with other
strategic territories in Southeast Asia could facilitate Japanese exports. It is believed that
0
1000
2000
3000
4000
5000
6000
0,0
0,5
1,0
1,5
2,0
2,5
1912 1915 1925 1929 1932 1938
Machinery Production (Millons yens)
Index Export ratio over production Machinery(1912=100)
Index Domestic Consumption ratio over productionMachinery (1912=100)
Index Import ratio over production Machinery(1912=100)
Colonial
91%Informal Empire
2%
Rest of the
World…
Machinary Exports 1938
15
those mechanisms explained in Section 2 helped to reduce information costs and
uncertainty. The reduction of information costs permitted a rise in exports, especially of
new products (at every skill-level), which, due to their novelty, are the ones suffering
larger information-related impediments to trade (see Martinicus et al. (2010)). In that
sense, the number of Japanese diplomats operating abroad is a suitable proxy for
measuring transnational commercial and investment networks and cultural diplomacy.
Figure 6 shows that after WWI the number of diplomats abroad and the number of new
products exported by Japan to countries within the future conquests followed the same
tendency. This fact suggests that Japanese interests in Southeast Asia worked mainly to
facilitate the introduction of newly developed Japanese products in those countries.23
Figure 6: Number of Japanese Diplomats and Products Exported Towards Japanese
Future Conquests (1912-1938).
Source: The number of Japan’s Diplomats abroad comes from Almanach de Gotha (various years, see
Section 4.2, Appendix A ) and the number of goods exported from Annual Returns of Trade of the Empire
of Japan (various years).
Previous evidence on Japanese export determinants seems to demonstrate that,
although improvements in productivity might be behind Japanese export performance
before the Great Depression, they do not explain most of the great increase in industrial
exports during the 1930s. This increase was strongly biased towards Japanese colonies
and was also directed towards territories occupied in the future, suggesting a greater
influence on export expansion by imperial mechanisms. Furthermore, the provisional
evidence shows that colonial mechanisms accelerated exports of high-skilled goods, for
which Japan did not have a comparative advantage. Additionally, links with future
dominions, which consisted mainly of migration, investment, and the creation of
23 Either the number of Japan’s diplomats abroad or the foreign diplomats in Japan is offered in Table 3 and
Table 4 in Section 4.2, Appendix A.
0
100
200
300
0
5
10
15
1912 1915 1925 1929 1932 1938
Nº Goods exported to Future ConquestsNº of diplomatics Future Conquests
16
transnational networks through cultural diplomacy in order to increase commercial
influence in Southeast Asia, permitted the introduction of new products. After presenting
our new data set in the next section, we will provide empirical confirmation of the
imperial power hypothesis using granular data on bilateral exports with a marginal
gravitational model.
4. The New Data Set
4.1 Trade Data.
As mentioned in the introduction, one of the principal contributions of this paper
is the construction of a granular data set that provides a high-resolution image of the value
of Japanese exports for each year and the geographical destination of exports by product
at the same level of disaggregation as official Japanese trade statistics. We digitised
information found in different volumes (1912, 1915, 1925, 1929, 1932, and 1938) of the
Annual Return of the Foreign Trade of the Empire of Japan. Official statistics collected
and published annually by the Japanese Bureau of Customs appear disaggregated at
product level and show quantities and values exported from Japan by destination.24
According to our data set, Japan moved from exporting 512 different products to 37
different countries in 1912, to exporting 1135 goods to 117 countries in 1938. On one
hand, the increase in the number of countries is related with more precise records of
destination, but also with a greater geographical diversification of Japanese exports over
the period. On the other hand, we suspect that the increase in the number of goods
exported might be related neither with real changes in the diversification of Japanese
industrial capacity, nor with changes in market characteristics, but simply with a change
in the classification rules employed in the official statistics. (For an extended discussion
of this point related with the extensive margins of trade specification, see Online
Appendix B.)
The main drawback of the Annual Return as a source is that it does not include
data on Japanese exports towards its main colonies, Korea and Taiwan. (Although,
24 We discuss the accuracy of the Japan trade statistics in Section 1, Appendix A. Although, we assume
Japan used quite accurate declared values for exports, we assume some inaccuracy on country real
destination records for goods moving through Hong Kong and Singapore. Nevertheless, we have evidence
that the Hong Kong share of total China imports declined from around 40% at the turn of the century to
less than 20% in the 1920s, and less than 5% in the 1930s. This implies that inaccuracy of Japan’s direction
of trade records due to transit is more limited for the interwar years than for previous periods (see Figure 5
in Keller et al. (2011)).
17
thankfully, they account for exports towards the Kwantung Leased Territory and
Manchuria, both before and after being occupied by Japan.) In order to solve this relevant
inaccuracy in the statistical records, and believing that the study of Japanese export
patterns towards its colonies is vital for our research interests, we used the information
on Japanese exports to Taiwan from the Returns of the Trade of Taiwan for Forty Years
(1896-1935) and the annual Return of Trade of Taiwan (Formosa) (1936-1942), both
published by the Department of Finance of the Government of Taiwan. These returns
include Taiwanese imports from Japan (or, equivalently, Japanese exports to Taiwan) at
product level.25 For the case of Korea, the Annual Returns of the Foreign Trade of the
Empire of Japan in their volumes from 1914-1920 include an appendix with data on
Japanese exports to Korea at the product level. For the rest of the period, Japanese exports
to Korea have been estimated by the Japanese Long Term Economic Statistics (LTES)
database. 26
The data set that we constructed includes the value of Japanese exports (in current
yen) of each product exported to every country. Each different product exported has been
categorised using the Standard International Trade Classification (SITC) Revision 2 to a
5 digit-level disaggregation to achieve a proper and precise product classification. This
categorisation also permits us to distinguish new export products from those exports that
are simply different varieties of the same product.27 The different goods have also been
divided according to skill-intensity using Board of Trade (1905), following Tena-Junguito
(2010). Skill-intensity is constructed based on wages earned by workers in each sector
25 The level of disaggregation included in this data set is lower than that of the main statistics after 1915,
so it would underestimate Taiwan’s extensive margins of trade. We have assumed that the number of
products exported to Taiwan after 1915 follows the same rate of increase as that of the Kwangtang Leased
Territory—the only Japanese colony for which there is full information throughout the period available in
the Annual Returns of Trade of the Empire of Japan. The Kwangtang Leased Territory was the colony that
received the greatest number of Japanese goods, which is why it could be considered an upper-bound for
calculating the evolution of the extensive margin to Taiwan. See an extended discussion in Section 1.2,
Appendix A.
26 For 1912 and 1915, there is full disaggregation of exports at the product level. For the period 1925-1938,
LTES includes data on Japanese exports to Korea in five different sectors that match our skill distribution.
Thus, the number of goods at every skill-level exported by Japan will be equal to the corresponding 1915
figure assumed to evolve along the same path as that of Kwantung, which shared a similar economic
structure to Korea. See discussion in Section 1.2, Appendix A.
27 For that purpose, products sharing the same 3-digit classification will compose a single product category
or sector. In Appendix B, we offer an extended discussion on the number of products recorded in the
statistical records and the methodology followed in the construction of sectors used to capture the Extensive
Margins of Trade.
18
and permits us to differentiate between high-skilled (Skill-intensity higher than 10),
medium-skilled (between 9 and 10), low-skilled manufactures (between 5 and 9), and
primary products (less than 5 skill-intensity).28
4.2 GDP and Productivity
Information on GDP and GDP per capita is taken from the latest version of the
Maddison Project and refers to GDP in 1990 GK dollars or real GDP (see Bolt and van
Zanden (2014)). For missing countries we took data on real GDP for 1950 and translated
it to the interwar years by assuming the same evolution as constant exports found at
Federico and Tena-Junguito (2018, 2019).
In addition, we estimated GDP levels and other statistics for Japanese colonies of
Kwantung Leased Territory and Manchukuo (Manchuria). Population data has been
obtained from Kang (1981), and the corresponding GDP is calculated as the proportion
of Japanese GDP equal to the percentage each population represents over the Japanese
total (the exercise for Manchuria is the same except China is used as the reference because
it was part of China until 1932). We considered Kwantung’s GDP to follow the same rate
of growth as Japan’s (because it was a Japanese colony), whereas we considered
Manchuria’s GDP to follow the same tendency as China’s until 1932. Afterwards, we
assumed Manchuria’s GDP to grow at the same rate as Manchurian exports taken from
Federico & Tena (2018). 29
GDP information is complemented by data on total hours worked, found in
Huberman and Mins (2007) and Maddison (2007 and 1995), to construct non-agrarian
GDP per hour worked as a measure of productivity for 27 partner countries, which
represent 92% of Japanese exports. The real wages index for capturing productivity
evolution for 30 partner countries comes from Williamson (1995), while population
density has been obtained from League of Nations data. Price indexes for deflating GDP
measures are from Mitchell (2003).30
Finally, one key part of this analysis is the study of Japanese productivity and the
study’s determination of exports by skill-level. In order to disentangle this relationship,
28 In Appendix C, we offer an extended discussion of the general division of total exports by skill-intensity.
29 For the estimation of Japanese colonies and other East Asian territories, see Section 2, Appendix C.
30 See Appendix A for an extended discussion on GDP per hour worked differentials (Section 3.1) and on
real wages and productivity differentials by country and sector (Section 3.2).
19
we employed data on Japanese manufacturing output in different sectors and the
respective employment per sector. We used data derived from total hours according to the
Long Term Economic Statistics of Japan (LTES) database from Hitotsubashi University
to approximate hours worked at every sector.
4.3 Trade Costs
Distance between countries is the most common variable to measure transport
costs in the literature on gravity, but this variable is time and product invariant. This paper
makes an effort to overcome the implications of this limited assumption on the measure
of transport cost. So we use a data set of estimated freight factors (cost of transport per
ton/product price per ton) between Japan and each partner country, for every single
product exported by Japan for each of our corresponding benchmark years. In our strategy
geographical distance between Japanese partners is combined with different freight rates,
routes, and differences in freight factors by composition of products.31
The information on the number of Japanese diplomats operating in each different
country has been obtained from the Almanach de Gotha (various years), which includes
a section with the names and country of origin of every diplomat operating in every
country. We have meticulously counted the number of Japanese diplomats that worked in
each country. Additional variables representing trade costs are average levels of tariff
protection, which come from Blattman et al. (2003), and exchange rate control, obtained
from various sources, including Bethell (1994), Eichengreen (1996), Meisel (1990), and
Reinhart and Rogoff (2002).32 Finally, bilateral nominal exchange rates with the Japanese
yen, determined as being the quantity of foreign currency per one yen, are found in
Federico and Tena (2018). In addition, we have used Federico and Tena-Junguito’s (2018,
2019) data set on nominal imports to capture the differences introduced by the
protectionist backlash between domestic and international demand, especially in the
1930s.
5. The Gravity Model
31 The alternative use of the conventional great-circle distance in our equations offers weak results similar
to the use of freight cost by products and regions. Apparently, this happens because freight factors of
manufactures are low. Additionally, international freight rates during the interwar years were volatile but
stagnant. See an extended discussion in Appendix E.
32 Tariffs between Japan and its colonies are assumed to be zero.
20
The traditional gravity model assumes that trade flows travel between two
countries attracted by two main forces. Countries’ economic size attracts trade flows,
whereas distance between countries discourages flows. The most basic form of the
Gravity model takes the following specification according to Burger et al. (2009):
𝐸𝑖𝑡 = 𝐾𝑀𝑖𝑡
𝛽1
𝐷𝑖𝑡𝛽2
EQ (1)
It is important to mention that sub-indexes “it” indicate that we used panel data to
estimate our equations. 𝑀𝑖𝑡 represents the economic size of country i in period t, while
𝐷𝑖𝑡 is equal to the distance to country i in period t. After taking logarithms on both sides,
the model can be estimated by OLS following Eichengreen & Irwin (1998). The equation
would take the following form in which the log of K is a constant that can be substituted
by more conventional 𝛽0:
𝑙𝑛𝐸𝑥𝑝𝑖𝑡 = 𝑙𝑛𝐾 + 𝛽1𝑀𝑖𝑡 − 𝛽2𝐷𝑖𝑡 + 𝜀𝑖𝑡 EQ (2)
Nevertheless, Santos Silva and Tenreyro (2006) argue that OLS is not a desirable
method of estimation for gravity models because there are many periods and many
countries to which exports are zero, so taking logarithms of the dependent variable would
not account for those zeros, causing the researcher to lose much valuable information.
They suggest employing Poisson Pseudo-Maximum Likelihood (PPML), an estimation
method that has recently become popular. Thanks to this model there is no need to take
logarithms of Japanese exports, thus keeping all information contained with regard to
those zeros. Furthermore, Gómez-Herrera (2013) mentions that PPML provides unbiased
estimates in cases of heteroscedasticity. The independent variables (with the exceptions
of dummies) are nevertheless logged for the sake of simplicity in their coefficients’
interpretations because this practice would permit us to interpret them as slopes.
This model, of course can be augmented by controlling for more variables, which
is a common practice in the literature. Two groups of variables are included, first, those
variables associated with comparative advantage, 𝐶𝐴𝑖𝑡 (related either with productivity,
relative wages, factor endowments, or similar demand structure), and second, those
variables associated with trade or transaction costs (transport cost, tariffs, exchange rates,
and empire connexions). The included Empire dummies would be part of the latter group,
but differentiated as potential reductions in transaction cost promoting trade.
In the comparative advantage group, first, we used relative productivity measures,
21
such as the Japanese GDP per hour worked relative to partner countries.33 Second, we
used a variable measuring relative real wage trends as ratio, so that a rise of the variable
reflects increases in Japanese wages relative to a trading partner. Next, a third type of
variable refers to differences in demand structure (proxied by GDP per capita
differentials) and captures whether Japanese exports responded to the pattern determined
by Linder (1961), in which trade between countries would be more intensive the more
similar their demand structures. This hypothesis has been successfully tested by Restrepo
and Tena (2016) for Latin America and by Meisner and Tang (2017) for Japan.34 Finally,
we checked whether Japanese trade followed a Hecksher-Ohlin pattern in which exports
are intensified when differences in factor endowments (population density in this case)
are high, as explained by Ohlin (1952), among others.
On the trade cost group, we include a variable called 𝑇𝐶𝑖𝑡 that is intended to
measure the effects of trade costs like average freight factors, average levels of protection
in partner countries, exchange rates with the Japanese yen, imposition of exchange
controls, or the number of Japanese diplomats abroad. Following our previous discussion
on the relevance of Empire, we add other transaction cost variables which are the key
ones in our analysis including: 𝐶𝑂𝐿𝑂𝑁𝑌𝑖𝑡 (a dummy taking the value 1 if a country is a
Japanese colony) that captures all those mechanisms by which Japanese imperial aims
fostered exports, and 𝐹𝑈𝑇𝑈𝑅𝐸 𝐶𝑂𝑁𝑄𝑈𝐸𝑆𝑇𝑆𝑖𝑡 that determines whether Japanese exports
were already biased towards countries occupied during WWII.
Finally, two additional variables are intended to capture the difference between
international and domestic demand occasioned by trade distortions and other path
dependence fundamentals. The first variable controls for Japanese partners’ imports and
is especially devoted to capturing the trade blocs and other trade diversion protectionist
measures that affected bilateral international demand above domestic GDP variations.
This variable would be especially relevant during the 1930s. The second one βt represents
33 The variable refers to GDP per hour worked excluding agriculture productivity. Additionally, for sector-
level analysis, we include Japanese product per hour on high- and low-skill manufactures found in LTES.
Then it is compared with the partner’s overall productivity following the same methodology as for real
wages. For further discussion regarding this variable, see Section 3, Appendix A.
34 The log of the absolute differences in GDP per capita between Japan and the corresponding partner
proxies for differences in demand structure. We have also tried to measure differential in demand structure
as the product of the bilateral income share (Yi)/(Yi+Yj)* (Yj)/(Yi+Yj), obtaining similar results because
taking logs reduces the differences between relative and absolute figures. Finally, several authors suggest
the use of differences in the degree of urbanisation or industrialisation as a proxy, these measures are very
controversial as representative of differential in demand structure, however, and in our case, it is very
difficult to collect representative data.
22
time dummies that permit us to control changes that are related with the passage of time
(like inflation or the First World War international disintegration) and εit are standard
errors clustered by country partner. Equation 3 is the one that we are going to estimate.
𝐸𝑥𝑝𝑖𝑡 = 𝛽0 + 𝛽1𝑙𝑛𝐺𝐷𝑃𝑖𝑡 + 𝛽2𝐶𝐴𝑖𝑡 + 𝛽3𝑇𝐶𝑖𝑡 + 𝛽4𝐶𝑂𝐿𝑂𝑁𝑌𝑖𝑡 +
𝛽5𝐹𝑈𝑇𝑈𝑅𝐸 𝐶𝑂𝑁𝑄𝑈𝐸𝑆𝑇𝑆𝑖𝑡+𝛽6𝐼𝑀𝑃𝑂𝑅𝑇𝑆𝑖𝑡 + 𝛽𝑡 + 𝜀𝑖𝑡 EQ
(3)
Another novelty of this article lies in the employment of a margins of trade
framework that allows us to determine by which channels the different independent
variables affected Japanese exports. In that sense, margins have been constructed in a
similar fashion as Meissner and Tang (2018), in which the extensive margin represents
the number of active sectors in which an exporter, Japan in this case, is exporting to each
country (i) and during each period (t). In our case, sectors are defined according to the 3-
digit level of the SITC Revision 2 description, in the sense that all products sharing the
first three digits will be collected within the same sector so we avoid accounting for
different varieties of existing products as though they were new goods (for an extended
discussion of this issue see Appendix B).35 By this methodology we observed that Japan
moved from exporting in 117 sectors to exporting in 164. The intensive margin, on the
other hand, represents the average nominal exports per sector exported to each country in
every period.36 Structural changes are usually related to an expansion in the extensive
margin, that is the number of different sectors exported. In this sense, Equation 4 displays
the extensive margins by employing the number of active sectors exported as the
dependent variable and Equation 5 uses exports per sector to country i in period t for
presenting intensive margins.
𝑁º𝑆𝐸𝐶𝑇𝑂𝑅𝑖𝑡 = 𝛽0 + 𝛽1𝑙𝑛𝐺𝐷𝑃𝑖𝑡 + 𝛽2𝐶𝐴𝑖𝑡 + 𝛽3𝑇𝐶𝑖𝑡 + 𝛽4𝐶𝑂𝐿𝑂𝑁𝑌𝑖𝑡 +
𝛽5𝐹𝑈𝑇𝑈𝑅𝐸 𝐶𝑂𝑁𝑄𝑈𝐸𝑆𝑇𝑆𝑖𝑡+𝛽6𝐼𝑀𝑃𝑂𝑅𝑇𝑆𝑖𝑡 + 𝛽𝑡 + 𝜀𝑖𝑡
35 We also employed a method similar to Huberman and Meisner (2017) in which the extensive margin
represents each single good exported by Japan. Results hold very similar. This method has been disregarded,
however, because it might bias our results as this method might show different varieties of existing products
as new products. See Appendix B.
36 IM=
𝑁𝑜𝑚𝑖𝑛𝑎𝑙 𝑒𝑥𝑝𝑜𝑟𝑡𝑠𝑖𝑡
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑆𝑒𝑐𝑡𝑜𝑟𝑠 𝑖𝑡
23
EQ(4)
𝐸𝑥𝑝𝑆𝐸𝐶𝑇𝑂𝑅𝑖𝑡 = 𝛽0 + 𝛽1𝑙𝑛𝐺𝐷𝑃𝑖𝑡 + 𝛽2𝐶𝐴𝑖𝑡 + 𝛽3𝑇𝐶𝑖𝑡 + 𝛽4𝐶𝑂𝐿𝑂𝑁𝑌𝑖𝑡 +
𝛽5𝐹𝑈𝑇𝑈𝑅𝐸 𝐶𝑂𝑁𝑄𝑈𝐸𝑆𝑇𝑆𝑖𝑡+𝛽6𝐼𝑀𝑃𝑂𝑅𝑇𝑆𝑖𝑡 + 𝛽𝑡 + 𝜀𝑖𝑡 EQ(5)
Finally, the model might experience some modifications because the dependent
variable could also be changed in order to determine how the variables of interest affected
Japanese exports by different skills products (high- and low-skills manufactures and
primary products) or what were the main determinants of exports to a specific region.
6. Results
6.1 Presentation of Results
The descriptive statistics presented in Section 3 showed that Japanese exports
increased significantly between 1912 and 1938, especially during the 1930s. These data
suggested also that, although Japanese improvements in productivity contributed to the
expansion of exports (either by different product varieties or more diverse destinations),
especially prior to 1929, they were not the main determinant of the structural change in
Japanese industrial exports. Evidence presented in previous sections suggested also that
among industrial exporters, Japan persisted with a comparative advantage in low-skilled
manufactures, such as textiles, although high-skilled exports increased substantially
during the 1930s. That evidence leads to our main hypothesis, developed in Section 3,
that imperial mechanisms, not productivity, were the main drivers that facilitated high-
skilled manufacturing exports, and links with future conquests fostered the introduction
of new Japanese products in the region.
This section will test the resilience of that hypothesis using a gravity model
strategy that takes advantage of the granular data that disentangles Japan’s trade margins.
First of all, we examine which factors determined Japan’s total exports during the whole
period. Independent variables (with the exception of dummies) are expressed in logs to
interpret them as elasticities.
Table 2 shows how the measures employed for approximating Japanese relative
productivity are significant when we do not control for Japanese colonies. However, upon
24
the inclusion of the colony dummy and other potential links with future conquests in
Southeast Asia, both productivity measures cease to be significant. This would reinforce
the trade in the shadow of power hypothesis that Japanese imperial connections might
compensate for lower productivity in facilitating overall exports. In that sense, the bias of
Japanese exports towards its colonies has the expected reaction. Furthermore, it is also
appreciable that Japan already had established certain economic relations that facilitated
exports towards those territories, which Japan subsequently occupied during WWII.37
According to the literature, another variable that may capture Japanese economic
interests in a territory before annexation is the number of diplomats in that territory, which
in this case is positive and significant, although its influence on Japanese exports vanishes
when we account for informal imperial links. This would suggest, as mentioned in Section
2, that the number of diplomats abroad and Japanese informal empire links worked
through the same mechanisms.38 Regarding the Japanese specialisation pattern, we can
appreciate how Japan tended to export more towards countries presenting greater
differences in demand structure (GDP per capita) and in population density (factor
endowments). This fact contradicts the Linder hypothesis for interwar Japan and places
its exports pattern closer to a Hecksher-Ohlin model. Assuming that Japan’s GDP per
capita was converging with rich countries and diverging with poor ones, this could
indicate that Japan specialised in exporting towards poorer countries. Finally, it is also
remarkable how among all the possible trade costs undermining Japanese exports, the
only significant ones are tariffs set by foreign countries and the relative appreciation of
the yen.
37 Larch et al. (2017) shows that the interpretation of dummy coefficients like Colony or Informal Empire
is the following: 𝑒2.612 − 1 = 1263% higher exports for colonies in comparison with non-colonial
territories and 𝑒0.842 − 1 = 132% for future occupied territories.
38 Table 11 in Appendix E shows that diplomats are still significant when only colonies are employed as a
control. The number of diplomats ceases to be a significant variable only when we account for informal
empire.
25
Table 2: Japanese Total Export Determinants
Gravity
Comp. Advantage
Trade Costs
Trade diversion
Empire
Note: Standard errors clustered by country. Time dummies included in the regression but excluded from
results display. *** p<0.01, ** p<0.05, *p<0.1
Once it becomes clear that imperial connections were more relevant than
productivity improvements on the determination of the direction of Japanese exports, it
is important to see which kinds of exports were affected by those mechanisms. Table 3
confirms our previous insights that Japanese comparative advantage determinants
especially affected low-skilled exports because relative productivity and real wage
increase had a bigger effect on these kinds of goods, suggesting that, according to market
forces, Japan was competitive producing low-skilled manufactures.
VARIABLES No Empire Empire EXPECTED
SIGN
GDP 1.435*** 1.328*** +
(0.398) (0.428)
RELPRODUCTIVITY 0.444* 0.432 +
(0.267) (0.389)
RELWAGES 0.990*** 0.412 +
(0.347) (0.398)
POPDENSITYDIFF 0.587*** 0.532*** +
(0.213) (0.178)
GDPCAPABSDIFF 1.220** 1.505*** +
(0.502) (0.462)
FREIGHTS -0.175 0.00824 -
(0.227) (0.219)
EXCHCONTROL 0.583 -1.128 -
(0.593) (0.816)
EXCHRATE -0.525*** -0.457*** -
(0.113) (0.122)
TARIFF -0.992*** -0.329* -
(0.121) (0.169)
IMPORTS 0.102** 0.167 +
(0.0508) (0.151)
COLONY 2.612*** +
(0.762)
FUTURE CONQUESTS 0.842** +
(0.332)
DIPLOMATS 0.185** 0.104 +
(0.0788) (0.0932)
Constant -5.362 -7.741*
(5.375) (4.574)
Observations 674 674
R-squared 0.651 0.640
26
On the other hand, colonial and other political mechanisms played a large, positive
role in every export category. Furthermore, the results in Table 3 show that being a
Japanese colony was by far the main factor attracting high-skilled exports (its effect being
much greater than that of productivity), suggesting that colonies became new markets for
those Second Industrial Revolution goods in which Japan did not possess a comparative
advantage. This market for high-skill goods was possible due to the mechanisms
previously explained. Contrary to total exports, the ‘future conquest’ regions are more
relevant to explain export growth on low-skill intensity exports. This would mean that
informal power mechanisms established in Southeast Asia before the Second World War
apparently reinforced Japanese comparative advantage within the region. 39
Table 3: Japanese Export Determinants by Skill Level (1) (2)
VARIABLES High Skills Low Skills
GDP 0.373 0.714*** (0.311) (0.212)
REL PRODUCTIVITY HS 0.939***
(0.336)
REL PRODUCTIVITY LS
1.083*** (0.182)
RELWAGES HS 0.0432
(0.321)
RELWAGES LS
0.618** (0.303)
POPDENSITYDIFF 0.502** 0.586*** (0.202) (0.123)
GDPCAPABSDIFF 1.737*** 0.383 (0.565) (0.433)
FFHIGHSKILL -0.229
(0.224)
FFLOWSKILL
-0.259 (0.164)
EXCHCONTROL -1.094 -0.605 (0.726) (0.438)
EXCHRATE -0.597*** -0.544*** (0.194) (0.0920)
TARIFF HS 0.0497
(0.116)
TARIFF LS
-0.298*** (0.0982)
COLONY 4.076*** 3.096*** (0.578) (0.355)
FUTURE CONQUESTS 1.270*** 2.517*** (0.416) (0.446)
DIPLOMATS 0.0815 0.243** (0.122) (0.113)
IMPORTS 0.338* 0.160***
(0.177) (0.0404)
39 Table 7 in Section 2, Appendix C, shows a disaggregated regression for high- and low-end textiles. Both
display similar behavior and present little difference with the overall determinants of low-skill exports.
27
Constant 1.322 -1.342 (4.031) (4.668)
Observations 676 676
R-squared 0.550 0.630
Note: Standard errors are clustered by country. Time dummies are included in the regression but are
excluded from the results display. *** p<0.01, ** p<0.05, *p<0.1
Finally, Table 4 distinguishes export determinants according to their effects on
either the number of different sectors in which Japan exported (extensive margin) or the
value exported per sector (intensive margin). This process is also replicated at skill-level.
The results show that relative productivity did not affect any margin and that the Japanese
Empire especially favoured exports in new sectors, which is linked with industrial
development and innovation.
Table 4: Japanese Export Determinants According to Extensive and Intensive
Margins (1) (2) (3) (4) (5) (6)
VARIABLES Extensive
Total
Intensive
Total
Extensive
HS
Extensive
LS
Intensive
HS
Intensive
LS
GDP 0.361*** 0.262*** 0.283*** 0.194* 0.275*** 0.701***
(0.112) (0.0914) (0.101) (0.112) (0.101) (0.224)
REL PRODUCTIVITY -0.216 -0.226
(0.238) (0.204)
REL PRODUCTIVITY HS 0.125 0.0661
(0.108) (0.103)
REL PRODUCTIVITY LS -0.504** 1.015***
(0.247) (0.162)
RELWAGES 0.220 0.225
(0.195) (0.182)
RELWAGES HS -0.202 -0.365**
(0.161) (0.173)
RELWAGES LS 0.673*** 0.560*
(0.173) (0.313)
POPDENSITYDIFF 0.00190 -0.0166 -0.0255 -0.0261 -0.0404 0.496***
(0.0631) (0.0543) (0.0742) (0.0663) (0.0649) (0.113)
GDPCAPABSDIFF -0.00312 0.00267 0.155 0.0339 0.0832 0.251
(0.217) (0.179) (0.232) (0.217) (0.201) (0.377)
FREIGHTS -0.286** -0.192
(0.139) (0.124)
FFHIGHSKILL -0.422*** -
0.317***
(0.122) (0.122)
FFLOW SKILL -0.282* -0.291*
(0.152) (0.175)
EXCHCONTROL -0.373* -0.554*** -0.699** -0.505** -
0.655***
-0.535
(0.224) (0.212) (0.325) (0.241) (0.223) (0.448)
EXCHRATE -0.0776 -0.0167 -0.101** 0.0136 -0.0220 -
0.503***
28
(0.0506) (0.0405) (0.0515) (0.0454) (0.0497) (0.0879)
TARIFF -0.160* -0.151**
(0.0943) (0.0747)
TARIFF HS 0.0122 -0.0145
(0.0723) (0.0637)
TARIFF LS 0.000150 -
0.327***
(0.0887) (0.0970)
COLONY 0.754* 0.400 0.788*** 0.437 0.498* 2.957***
(0.400) (0.344) (0.295) (0.370) (0.302) (0.354)
FUTURE CONQUESTS 0.776*** 0.494** 0.511 0.467 0.393 2.535***
(0.228) (0.194) (0.382) (0.430) (0.350) (0.461)
DIPLOMATS 0.100 0.136** 0.130* 0.116 0.140* 0.209
(0.0752) (0.0647) (0.0733) (0.0715) (0.0739) (0.144)
IMPORTS 0.0352 0.0229 0.0248 0.0269 0.0187 0.143***
(0.0453) (0.0301) (0.0406) (0.0412) (0.0349) (0.0289)
Constant -0.490 -2.764** -1.152 -1.188 -
3.680***
-2.951
(1.522) (1.259) (1.384) (1.308) (1.305) (2.355)
Observations 674 674 676 676 676 676
R-squared 0.488 0.506 0.506 0.411 0.471 0.755
Note: Standard errors are clustered by country. Time dummies are included in the regression, but are
excluded from the results display. *** p<0.01, ** p<0.05, *p<0.1
The most important conclusion we reach disaggregating this study by skill-level
is that high-skilled exports are more sensitive to the extensive margin than low-skill
exports. The greater connection between fixed costs and increases in new export products
in the high-skilled sector helps to understand that greater sensitivity. We can also
appreciate how improvements in Japanese productivity and comparative advantage only
increased the intensive margin for low-skilled products, which is apparently the sector
less linked to new technological products or innovation. On the other hand, being
colonized by Japan was the main driver of new extensive exports of high-skilled
manufactures, suggesting that those newer and more sophisticated products were not
internationally competitive and were preferred to be traded under the shadow of power in
Japanese colonies. This situation does not exclude the fact that both colonies and future
conquests also facilitated low-skilled exports at the intensive margin in a relevant way.
This section confirms some of the hypotheses offered in the descriptive analysis.
In contrast to the influence exercised by colonial mechanisms, Japanese productivity
played a marginal role in determining total Japanese exports during the interwar years.
Additionally, it seems that prior to Japanese occupation there was an increase in economic
and commercial exchange with Southeast Asia. We have also found that Japanese
comparative advantage and increases in productivity were more related to low-skill than
to high-skill manufactures, which were directed towards poor countries. Most
29
importantly, we found that mainly Japanese colonies facilitated exports of high-skilled
goods, whereas informal empire mechanisms reinforced the existing comparative
advantage in the export of low-skilled goods within the region. Finally, we have also
determined that colonial and other economic links built prior to the annexation of
Southeast Asian territories permitted Japan to introduce new kinds of products not
previously exported to the region. In other words, it seems that the Japanese shadow of
power permitted the country to complete its industrialisation process by developing an
imperial market for its more sophisticated manufactures.
6.2 Robustness Checks
In order to check whether our results are robust, we tested whether the previous
conclusions still held when we employed different estimation methods and limited the
potential endogeneity between the key variables. In that sense, Table 5 shows our main
results estimated by a Tobit estimator, employing OLS, and a later panel estimation using
Random Effects to finish with some checks on endogeneity.
Our results generally hold true to our hypothesis that Japanese relative
productivity did not affect Japanese exports regardless of the different estimation methods
employed. First, being a Japanese colony was always the most important determinant of
Japanese exports. Second, being part of Japan’s future dominions was a significant
additional attractor of Japanese exports in all estimation methods. Finally, trade costs and
comparative advantage worked in the same direction as before.
Table 5: Japanese Export Determinants Using Alternative Estimation
Methods
(1) (2) (3) VARIABLES Tobit OLS Panel Random
Effects
ExpectedSign
GDP 0.769*** 0.769*** 0.737*** +
(0.152) (0.243) (0.198)
REL PRODUCTIVITY 0.525 0.525 0.134 +
(0.343) (0.587) (0.449)
RELWAGES 0.488 0.488 0.535 +
(0.338) (0.502) (0.447)
POPDENSITYDIFF -0.146 -0.145 -0.114 +
(0.140) (0.207) (0.160)
30
GDPCAPABSDIFF 0.713** 0.712 0.874*** +
(0.288) (0.525) (0.286)
FREIGHTS -0.498* -0.499 -1.044*** -
(0.269) (0.339) (0.210)
EXCHCONTROL 0.116 0.116 -0.232 -
(0.855) (0.540) (0.609)
EXCHANGERATE -0.870*** -0.870*** -0.639*** -
(0.111) (0.192) (0.157)
TARIFF -0.534*** -0.535*** 0.0789 -
(0.147) (0.200) (0.157)
COLONY 2.642*** 2.637*** 3.020*** +
(0.610) (0.825) (0.784)
FUTURE CONQUESTS 1.686*** 1.683*** 1.405** +
(0.331) (0.533) (0.587)
DIPLOMATS 0.453*** 0.452** 0.304** +
(0.134) (0.205) (0.152)
IMPORTS -0.0334 -0.0334 0.0143 +
(0.0327) (0.0525) (0.0472)
Constant 7.009*** 7.017** 5.496***
(1.876) (3.532) (2.096)
Observations 314 314 314
R-squared 0.1532 0.523 0.3607
Note: Standard errors are clustered by country. Time dummies are included in the regression, but are
excluded from the results display. *** p<0.01, ** p<0.05, *p<0.1
We also checked the determinants of exports in the formal and informal empire
across benchmarks years to detect the different influences of change in imperial policy
throughout the period. Table 6 shows that Japanese productivity was significant only in
1932, but not in 1938, when regional exports reached a peak. On the other hand, being a
colony strongly affected Japanese exports in a positive way after WWI. The importance
of this factor being much greater in 1938, when Japanese trade policy was clearly directed
towards its dominions.
Table 6: Japanese Export Determinants Towards Countries Inside Its Formal and
31
Informal Empire
VARIABLES IMPERIAL EXPORTS EXP SIGN
GDP 0.133 +
(0.248)
RELPRODUCTIVITY 1915 3.450 +
(2.299)
RELPRODUCTIVITY 1925 2.344* +
(1.253)
RELPRODUCTIVITY 1929 1.542* +
(0.816)
RELPRODUCTIVITY 1932 2.943** +
(1.260)
RELPRODUCTIVITY 1938 2.050 +
(1.297)
RELWAGES 0.599** +
(0.304)
POPDENSITYDIFF 0.906*** +
(0.319)
GDPCAPABSDIFF -0.390 +
(0.535)
FREIGHTS -0.841 -
(0.551)
EXCHCONTROL 0.248 -
(0.585)
EXCHRATE -2.618*** -
(0.903)
COLONY 1915 0.374 +
(0.643)
COLONY 1925 1.764*** +
(0.523)
COLONY 1929 1.177*** +
(0.412)
COLONY 1932 2.152*** +
(0.590)
COLONY 1938
3.835*** +
(0.582)
IMPORTS 0.158***
(0.0443)
DIPLOMATS 0.0969 +
(0.162)
Constant 6.022
(4.486)
Observations 96
R-squared 0.900
Note: Standard errors are clustered by country. Time dummies are included in the regression, but are
excluded from the results display.*** p<0.01, ** p<0.05, *p<0.1
32
Finally, one important drawback of our results could be the presence of
endogeneity in different ways. First, there is a possibility that our results suffer from
endogeneity by simultaneity in the sense that a military occupation and the creation of
colonies might increase Japanese exports, but there is also a possibility that previous
economic and commercial links might motivate a military intervention. This fact could
explain why the impact of Japanese colonies on exports found in our results is so large.
Fortunately, the case of Manchuria, which for most of the period was not a Japanese
colony, and which was occupied in 1931, might be useful for controlling for such
endogeneity.
For that purpose, Japanese exports to Manchuria before and after military
annexation are exploited to run a diff-in-diff estimation. In this case, Manchuria will be
considered the treated group, as it was invaded by Japan in 1931. This event will be
considered the intervention. Thus, the periods before 1932 are defined as the pre-
treatment period and that afterward as the post-treatment period.40 The control group will
be all of those countries inside the region, whose performances as recipients of Japanese
exports were similar to that of Manchuria before the intervention. Additionally, Japanese
intervention in Manchuria will be considered as an exogenous event in the sense that,
controlling for economic and diplomatic characteristics and transport costs, every country
inside Japan’s sphere of influence could have been colonised instead of Manchuria.41 We
used the following diff-in-diff equation to estimate the effects of Japanese military
intervention on Japanese exports towards Manchuria.
𝐸𝑥𝑝𝑖𝑡 = 𝛽0 + 𝛽1𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖𝑡 + 𝛽2𝑃𝑜𝑠𝑡1932𝑖𝑡 + 𝛽3𝑇𝑟𝑒𝑎𝑡𝑒𝑑 ∗ 𝑃𝑜𝑠𝑡1932𝑖𝑡 +
𝛽4𝑋𝑖𝑡 + 𝛽𝑡 + 𝜀𝑖𝑡 EQ (6)
In this equation 𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖𝑡 is a dummy that represents Manchuria, which is the
only country in the sample that became a consolidated colony during the interwar years.
(Part of China was also considered a Japanese colony by 1938, but it was not consolidated
at all.) 𝑃𝑜𝑠𝑡1932𝑖𝑡 is a dummy which equals 1 in 1938 (the post-intervention period).
40 The military intervention took place in late-1931, but the colony was not established until early-1932.
For that reason, it seems unlikely that the effects of colonisation would be evident on 1932 exports. That is
why 1938 is considered as the only post treatment period.
41 In reality Japan was more likely to intervene in China or Manchuria for political and strategic reasons
than in any of the other countries in the region, but theoretically the assumption is plausible.
33
Finally, 𝑇𝑟𝑒𝑎𝑡𝑒𝑑 ∗ 𝑃𝑜𝑠𝑡1932𝑖𝑡 is the diff-in-diff variable that represents the true effect
of Japanese colonisation on exports to Manchuria.42 𝑋𝑖𝑡 represents the controls previously
explained, including a control of countries that were part of Japanese dominions during
WWII, whereas 𝛽𝑡 are time-specific fixed effects. We used this equation to estimate the
effects for countries inside the Japanese sphere of influence.
The results are given in Table 7. First, we note that Japanese total and high-skilled
exports to Manchuria were lower than to the rest of the region for the whole period.
Second, the results also show that both kinds of exports are higher after 1938 than before
(although not significantly). Finally, and more importantly, the diff-in-diff variable is
positive and significant, meaning that controlling for other variables, exports to
Manchuria rose after occupation relative to the rest of the region. This suggests that once
colonization endogeneity is avoided, the Japanese imperial export driver is robust as the
explanation of export expansion. Furthermore, the effect of military occupation on high-
skill exports is still very relevant and significant. This fact reinforces our premise that
colonisation was the main driver for Japanese introduction of Second Industrial
Revolution goods into East Asia.
Table 7: Diff-in-Diff Estimation for Japanese Exports Before and After Colonisation
(1) (2)
VARIABLES Total Exports High-Skilled
TREATMENT -1.874*** -3.307***
(0.420) (0.404)
POST1932 0.682 0.529
(0.501) (0.472)
TREATMENTPOST1932 1.331*** 3.617***
(0.271) (0.311)
Observations 96 96
R-squared 0.938 0.968
Note: Standard errors are clustered by country. Time dummies and the rest of controls included on previous
42 Effect of intervention in the treated country=(𝛽0 + 𝛽1 + 𝛽2 + 𝛽3) − (𝛽0 + 𝛽1) = 𝛽2 + 𝛽3
Effect of (non)intervention in the control group= (𝛽0 + 𝛽2) − 𝛽0 = 𝛽2
Total effect of military intervention= (𝛽2 + 𝛽3) − 𝛽2 = 𝛽3
34
regressions are accounted in the calculations, but are excluded from the results display.*** p<0.01,**
p<0.05, *p<0.1
Another potential source of endogeneity comes from the impact that export
growth may have had on productivity growth, transport cost, or commercial diplomacy.
We offer some reverted explicative equations in Appendix E. Table 9, shows that exports
did not affect Japanese productivity during the interwar years (or did so only in a weak
way) and the potential that trade costs and exports were bidirectional because increasing
exports might have further reduced trade costs between countries. Appendix E shows that
this bidirectionality was real for Japanese exports and its diplomats abroad. To avoid the
possibility that this phenomenon might spoil the effect of trade determinants, we lagged
the independent variables. The results in Table 8 show that the main conclusions still hold
after avoiding the potential for current exports to affect previous trade determinants. We
can also see that our productivity and trade costs measurements (including empire) have
a robust effect on exports, which persists at least for one period ahead.
Table 8: Japanese Exports Determinants (Lagged Independent Variables)
(1) VARIABLES EXP
GDP 0.977***
(0.300)
RELPRODUCTIVITY 0.544*
(0.292)
RELWAGES 0.125
(0.140)
GDPCAPABSDIFF 1.035**
(0.408)
POPDENSITYDIFF 0.601***
(0.125)
FREIGHTS -0.219
(0.177)
TARIFFS -0.338**
(0.140)
EXCHCONTROL -1.757
(1.097)
EXCHRATE -0.457***
(0.109)
COLONY 2.638***
(0.706)
FUTURE CONQUESTS 0.832***
(0.305)
DIPLOMATS 0.375***
(0.109)
IMPORTS 0.156*
(0.0861)
Constant -4.937
(3.044)
35
Observations 673
R-squared 0.741
Note: Standard errors are clustered by country. Time dummies and the rest of controls included on previous
regressions are accounted in the calculations, but are excluded from the results display.*** p<0.01,**
p<0.05, *p<0.1
7. Conclusions
Trade blocs in the interwar years were used as instruments of imperial power to
foster exports and as a substitute for productivity in encouraging industrial production.
We have assumed here that the imperial shadow of power in interwar Japan predicted the
regional trade bloc formation of ‘The Greater East Asia Co-Prosperity Sphere’. Although
this bloc was achieved only by means of military intervention in the 1940s, it began by
working through certain economic interests that fostered regional exports in East and
Southeast Asia. Certainly, this research calls for a deeper study of the specific trade
finance connections between banks and large companies (Zaibatsu), which, in concert
with strategic military measures, encouraged bilateral trade with territories belonging to
the present and future colonial empire.
During the interwar years, and especially in the 1930s, as a reaction to the strong
disintegration of the international economy, Japan activated a complex industrialisation
policy to foster industrial production in the Second Industrial Revolution sectors of steel,
machinery, and mechanical engineering. This policy was based on an overlapping of
protectionist import substitution and imperial policies to stimulate exports in its region of
influence. That new policy reshaped previous Japanese export specialisation patterns,
which had been based on exporting low-skilled manufactures to poor countries. After the
First World War, Japan built a strong comparative advantage in the region, exporting low-
skilled manufactures to countries with different demand preferences and climatic
characteristics. Later, old colonies were used to foster the demand of new railway
infrastructures and machinery. Additionally, previous Japanese investment and
distribution channels, built around polities and countries that were subsequently occupied
during WWII, reinforced comparative advantage in low-skilled exports by facilitating an
expansion of the number of products exported to those territories. We proved that this
expansion occurred even before annexation. In that sense, we can identify a dual
relationship between Japanese economic imperialism and military imperialism. In non-
colonial territories of China and Southeast Asia, economic imperialism anticipated
military occupation whereas in the colonial territories, military conquest (also preceded
36
by economic penetration) exacerbated commercial exchanges in high-skilled
manufactures and fostered deeper industrialisation in Japan.
We are aware that a potential endogeneity problem exists that relates import
dependence or bilateral trade connections with military interventions (see Bonfani and
O’Rourke (2014)). Nevertheless, we believe that these results reasonably prove the
colonial trade bias mechanism used by imperial institutions, inversely related to
productivity, in order to support Japanese industrialisation in the first half of the twentieth
century. Apparently, the market potential for high-skilled manufactures was very limited.
Colonial intervention, therefore, was used to expand the market of the most intensive
human capital sector of Japanese industrialisation. The implicit counterfactual hypothesis
would be that without imperial intervention in the region, Japan would not have expanded
its high-skilled exports and would not have exported such a variety of new products. In
other words, Japan’s industrialisation process would have been much less pronounced.
8. Bibliography
Aziz, M. A. (2012). Japan’s Colonialism and Indonesia, Vol. I. Berlin: Springer Science
& Business Media.
Almanach de Gotha. Annuaire généalogique, diplomatique et statistique. Gotha: Perthes.
various years.
Bassino, J. P., and van der Eng, P. (2002). ‘Economic divergence in East Asia: new
benchmark estimates of levels of wages and GDP, 1913-1970’. Thirteenth IEHA
Congress, Buenos Aires, 22-26 July 2002.
Bethell, L. (ed.) (1994). The Cambridge History of Latin America, Vol. VI: 1930 to the
Present, Part 1. Cambridge University Press.
Bix, H. (1972). ‘Japanese Imperialism and the Manchurian Economy, 1900-31’. The
China Quarterly, 51: 425-443, available at http://www.jstor.org/stable/652483.
Blattman, C., Clemens, M. A., and Williamson, J. G. (2003) ‘Who Protected and Why?
Tariffs the World Around 1870-1938’. Harvard Institute of Economic Research Working
Papers 2010.
Board of Trade (1905). ‘The Comparative Incidence of Foreign and Colonial Import
Tariffs on the Principal Classes of Manufactures Exported from the United Kingdom’.
British Parliamentary Papers, No. 84: 280-320.
Bolt, J. and van Zanden, J. L. (2014). ‘The Maddison Project: Collaborative Research on
Historical National Accounts’. The Economic History Review, 67 (3): 627–651, working
paper.
37
Bonfatti, R. and O’Rourke, K. (2014). ‘Growth, Import Dependence and War’. University
of Oxford Discussion Papers in Economic and Social History. No. 132, July 2014.
Broadberry, S. N., Fukao, K., and Zammit, N. (2015). ‘How Did Japan Catch-up on the
West? A Sectoral Analysis of Anglo-Japanese Productivity Differences, 1885-2000’.
Working Paper Series 231, Competitive Advantage in the Global Economy (CAGE).
Burger, M., van Oort, F., and Linders, G. J. (2009). ‘On the Specification of the Gravity
Model of Trade: Zeros, Excess Zeros and Zero-inflated Estimation’. Spatial Economic
Analysis, 4 (2): 167-190.
Cain, P. J. and Hopkins, A. G. (1980). ‘The Political Economy of British Expansion
Overseas, 1750–1914’. The Economic History Review, 33 (4): 463-490.
Chase, K. A. (2005). Trading Blocs: States, Firms, and Regions in the World Economy.
Ann Arbor: University Michigan Press.
Chenery, H. B., Shishido, S., & Watanabe, T. (1962). ‘The Pattern of Japanese Growth,
1914-1954’. Econometrica, 30: 98-139.
De Bromhead, A., Fernihough, A., Lampe, M. and O'Rourke, K. H., 2017. ‘When Britain
Turned Inward: Protection and the Shift Towards Empire in Interwar Britain’, QUCEH
Working Paper Series 2017-02, Queen's University Belfast, Queen's University Centre
for Economic History.
Duus, P. (1998). Modern Japan. Boston: Houghton Mifflin.
Eckstein, A., Chao, K., and Chang, J. (1974). ‘The Economic Development of Manchuria:
The Rise of a Frontier Economy’. The Journal of Economic History, 34 (1): 239-264.
Eichengreen, B. (1996). Golden Fetters: The Gold Standard and the Great Depression,
1919-1939. Oxford University Press.
Eichengreen, B. and Irwin, D. A. (1995). ‘Trade Blocs, Currency Blocs and the
Reorientation of World Trade in the 1930s’. Journal of International Economics, 38 (1-
2): 1-24.
Eichengreen, B. and Irwin, D. A. (1998). ‘The Role of History in Bilateral Trade Flows’.
In Frankel, J. A. (ed.), The Regionalization of the World Economy. University of Chicago
Press, 33-62.
Ellinger, B. and Ellinger, H. (1930). ‘Japanese Competition in the Cotton Trade’. Journal
of the Royal Statistical Society, 93: 185-232.
Encarnation, D. J. (ed.) (1999). Japanese Multinationals in Asia: Regional Operations in
Comparative Perspective. Oxford University Press.
Federico, G. and Tena-Junguito, A. (2016). ‘World Trade, 1800-1938: A New Data-set’.
EHES Working Papers in Economic History, No. 93.
38
Federico, G. and Tena-Junguito, A. (2018). Federico-Tena World Trade Historical
Database, available at http://www.uc3m.es/tradehist_db.
Federico, G. and Tena-Junguito, A. (2019). ‘World Trade, 1800-1938: A New
Synthesis’. Revista de Historia Económica-Journal of Iberian and Latin America
Economic History, 37 (1): 9-41.
Findlay, R. and O’Rourke, K. (2007). Power and Plenty. Princeton, NJ: Princeton
University Press.
Fisher, C. (1950). ‘The Expansion of Japan: A Study in Oriental Geopolitics: Part II.
The Greater East Asia Co-Prosperity Sphere’. The Geographical Journal, 115 (4/6):
179-193, available at doi:10.2307/1790152
Fletcher, W. M. (1989). The Japanese Business Community and National Trade Polity,
1920-1940. Chapel Hill: University of North Carolina Press.
Francks, P. (2015). Japanese Economic Development: Theory and Practice. Abingdon:
Routledge
Gallagher, J. and Robinson, R. (1953). ‘The Imperialism of Free Trade’. The Economic
History Review, New Series, 6 (1): 1-15, available at doi:10.2307/2591017.
Garfinkel, M, Skaperdas, S., and Syropoulos, C. (2012). ‘Trade in the Shadow of
Power’. In Skaperdas, S., and Syropoulos, C. (eds.), Oxford Handbook on the
Economics of Peace and Conflict. Oxford University Press.
Gómez-Herrera, E. (2013). ‘Comparing Alternative Methods to Estimate Gravity
Models of Bilateral Trade’. Empirical Economics, 44 (3): 1087-1111.
Gross, S. (2016). Export Empire: German Soft Power in Southeastern Europe, 1890–
1945, Cambridge University Press.
He, Y. (2007). ‘History, Chinese Nationalism and the Emerging Sino–Japanese
Conflict’. Journal of Contemporary China, 16 (50): 1-24.
Howe, C. (1999). The Origins of Japanese Trade Supremacy: Development and
Technology in Asia from 1540 to the Pacific War. University of Chicago Press.
Huberman, M. and Minns, C. (2007). ‘The Times They are Not Changin’: Days and Hours
of Work in Old and New Worlds, 1870–2000’. Explorations in Economic History, 44 (4):
538-567.
Huberman, M., Meissner, C. M., and Oosterlinck, K. (2017). ‘Technology and Geography
in the Second Industrial Revolution: New Evidence from the Margins of Trade’. The
Journal of Economic History, 77 (1): 39-89.
Hummels, D., and Klenow, P. J. (2005). ‘The Variety and Quality of a Nation’s
Exports’. American Economic Review, 95 (3): 704-723.
39
Isserlis, L. (1938). ‘Tramp Shipping Cargoes, and Freights’. Journal of the Royal
Statistical Society, 101 (1): 53-146.
Iyenaga, T. (1912). ‘Japan in South Manchuria’. The Journal of Race Development, 2
(4): 373-398, available at doi: 10.2307/29737926.
Kang, C. (1981). ‘Demographic Development in Manchuria, 1924-1941’. 近代史研究所
集刊 (Bulletin of the Academia Sinica), 10: 351-375.
Ohkawa, K., Shinohara, M., and Umemura, M. (1967-89). Long-Term Economic
Statistics (LTES), Tokyo: Toyo Keizai Shinposha.
Koh Soo Jin, D. and Tanaka, K. (1984). “Japanese Competition in the Trade of Malaya
in the 1930s”. Southeast Asian Studies, (21) 4, March.
Kohli, A. (1994). ‘Where Do High Growth Political Economies Come From? The
Japanese Lineage of Korea's “Developmental State”’. World Development, 22 (9): 1269-
1293.
Kublin, H. (1959). ‘The Evolution of Japanese Colonialism’. Comparative Studies in
Society and History, 2 (1): 67-84.
Larch, M., Wanner, J., Yotov, Y., and Zylkin, T. (2017). ‘The Currency Union Effect: A
PPML Re-assessment with High-Dimensional Fixed Effects’. School of Economics
Working Paper Series 2017-7, LeBow College of Business, Drexel University.
League of Nations. (1927-1945). Statistical Year-book of the League of Nations. Geneva:
League of Nations Economic and Financial Section.
Linder, S. (1961). An Essay on Trade and Transformation. Uppsala: Almqvist & Wiksell.
Maddison, A. (1995). Monitoring the World Economy, 1820-1992. Paris: OECD
Development Centre.
Maddison, A. (2007). The World Economy: Two-in-One Edition (Vol. 1: A Millennial
Perspective, Vol. 2: Historical Statistics). Geneva: OECD Development Centre & New
Delhi: Academic Foundation.
Martincus, C. V., Estevadeordal, A., Gallo, A., and Luna, J. (2010). ‘Information Barriers,
Export Promotion Institutions, and the Extensive Margin of Trade’. Review of World
Economics, 146 (1): 91-111.
Matsusaka, Y. T. (2007). ‘The Japanese Empire’. In Tsutsui, M. (ed.), A Companion to
Japanese History. Malden: Blackwell, 224-40.
Meisel, A. (1954 - [et al.] (1990). El Banco de la República: antecedentes, evolución y
estructura. Bogotá: Banco de la República. Departamento editorial.
Meissner, C. M., and Tang, J. P. (2018). ‘Upstart Industrialization and Exports: Evidence
from Japan, 1880-1910’. The Journal of Economic History, 78 (4): 1068-1102.
40
Milward, A. S. (1981). ‘The Reichsmark Bloc and the International Economy’, in G.
Hirschfeld, G. And Kettenacker, L. (eds.), Der ‘Führerstaat’: Mythos und Realität:
Studien zur Struktur und Politik des Dritten Reiches. Stuttgart: Klett-Cotta.
Ministry of Finance. Annual Return of the Foreign Trade of the Empire of Japan. Tokyo:
Yushodo (1912, 1915, 1925, 1929, 1932, 1938).
Mitchell, B. R. (1971). Abstract of British Historical Statistics. Cambridge University
Press.
Mitchell, B. R. (ed.) (2003). International Historical Statistics: Africa, Asia & Oceania,
1750–2000. Basingstoke & New York: Palgrave Macmillan.
Moneta, C. (1959). ‘The Estimation of Transportation Costs in International Trade’.
Journal of Political Economy, 67 (1): 41-58.
Neal, L. (1979). ‘The Economics and Finance of Bilateral Clearing Agreements:
Germany, 1934-38’. The Economic History Review, Second Series, 32 (3): 391-404.
Ohlin, B. (1952). Interregional and International Trade, Vol. 39. Cambridge, Mass.:
Harvard University Press.
Osterhammel, J. (1986). ‘Semi-Colonialism and Informal Empire in Twentieth-Century
China: Towards a Framework of Analysis’. In Mommsen, W. J. (ed.), Imperialism and
After: Continuities and Discontinuities. London : Allen & Unwin, 290-314.
Petri, P. A. (1993). ‘The East Asian Trading Bloc: An Analytical History’. In Frankel, J.
and Kahler, M. (eds.), Regionalism and Rivalry: Japan and the United States in Pacific
Asia. University of Chicago Press, 21-52.
Post, P. (1996). ‘Characteristics of Japanese Entrepreneurship in the Pre-War
Indonesian Economy’. In Lindblad, J. T. (ed.), Historical Foundations of a National
Economy in Indonesia, 1890s-1990s, Amsterdam: Royal Netherlands Academy of Arts
and Sciences, 297-314.
Reinhart, C. and Rogoff, K. (2004). ‘The Modern History of Exchange Rate
Arrangements: A Reinterpretation’. The Quarterly Journal of Economics, 119 (1): 1-48.
Restrepo-Estrada, M. I. and Tena-Junguito, A. (2016). The Roots of Regional Trade in the
Americas 1870 to 1950. IFCS Working Papers in Economic History. WH 23304,
Universidad Carlos III de Madrid. Instituto Figuerola.
Ritschl, A. O. (2001). ‘Nazi Economic Imperialism and the Exploitation of the Small:
Evidence from Germany’s Secret Foreign Exchange Balances, 1938-1940’, Economic
History Review 54 (2): 324-45.
Royal Statistical Office. (1880). Statistical Abstract for the Several Colonial and Other
Pos-
Sessions of the United Kingdom (Statistical Abstract for the British Empire).
41
London: Eyre & Spottiswoode.
Sanderson, A. E. (1926). Ocean Freight Rates in United States Foreign Trade.
Washington: Government Printing Office [United States Bureau of Foreign and Domestic
Commerce].
Sanderson, A. E. (1940). Wartime Control of Ocean Freight Rates and Foreign Trade: A
World Survey. Washington: Government Printing Office.
Santos Silva, J. C. M., and Tenreyro, S. (2006). ‘The Log of Gravity’. The Review of
Economics and Statistics, 88 (4): 641-658.
Schuler, K. (2004). Tables of Modern Monetary History: Asia. Preliminary version, 29.
Shiraishi, T. and Shiraishi, S. S. (eds.) (1993). The Japanese in Colonial Southeast Asia,
Vol. 3. Ithaca, N.Y.: Cornell University Press.
Smitka, M. (ed.) (1998). Agricultural Growth and Japanese Economic Development, Vol.
7. New York & London: Garland.
Souza, G. B. (2004). The Survival of Empire: Portuguese Trade and Society in China and
the South China Sea, 1630-1754. Cambridge University Press.
Sugiyama, S. (2013). Japan's Industrialization in the World Economy: 1859-1899:
Export, Trade and Overseas Competition. London: Bloomsbury Academic.
Tena-Junguito, A. (2010). ‘Bairoch Revisited: Tariff Structure and Growth in the Late
Nineteenth Century’. European Review of Economic History, 14 (1): 111-143.
Wolcott, S. (1994). ‘The Perils of Lifetime Employment Systems: Productivity Advance
in the Indian and Japanese Textile Industries, 1920–1938’. The Journal of Economic
History, 54 (2): 307-324.
Williamson, J. G. (1995). ‘The Evolution of Global Labor Markets since 1830:
Background Evidence and Hypotheses’. Explorations in Economic History, 32 (2): 141-
196.
Williamson, J. A. (2002). ‘Is Protection Bad for Growth? Will Globalization Last?
Looking for Answers in History’. Thirteenth IEHA Congress, Buenos Aires, 22-26 July
2002.
Yasuba, Y. (1978). ‘Freight Rates and Productivity in Ocean Transportation for Japan,
1875-1943’. Explorations in Economic History, 15 (1): 11-39.
42
APPENDIX
APPENDIX A
Database Methods and Sources:
1 Trade data by product and country.
1.1.Japanese trade data
We have constructed a granular database of the value of bilateral Japanese exports
by product for different benchmark years at the same level of disaggregation as those
offered by the official Japanese trade statistics. We have digitised information found in
different volumes (1912, 1915, 1925, 1929, 1932 and 1938) of the Annual Return of the
Foreign Trade of the Empire of Japan. Official statistics collected and published annually
by the Japanese Bureau of Customs include a section that collects Japanese exports to
every country, disaggregated by single product. Japan used a quite accurate declared
values c.i.f system, for imports and f.o.b valuation for exports, and gold and silver are
excluded from trade statistics. Japan followed as far as possible the origin-destination
method of crediting trade by countries. However, one faces difficulties in getting
information on country of origin and destination for goods moving through Hong Kong
and Singapore.43 As we have evidence that Hong Kong’s share of total Chinese imports
reduced from around 40% at the turn of the century to less than 20% in the 1920’s and
less than 5% in the 1930´s, we assume that the inaccuracy of the Japanese direction of
trade records due to transit is more limited for the interwar years than for previous periods
(see Figure 5 in Keller et al (2011).44 According to our database, Japan moved from
exporting 512 different products to 37 different countries in 1912, to exporting 1135
goods to 117 countries in 1938.45 On one hand, more countries records is related with a
more precise record of destination, but also with a more extended geographical
diversification of Japanese exports. On the other hand, we suspect that the increment in
the number of goods exported is neither related with true changes in the Japanese
43 See Chapter 4 and Chapter 5 in Allen and Elly (1953). 44 By instance, Meissner and Tang (2018) do not make any adjustment to overcome Hong Kong and
Singapore transit trade problems on Japan geographical trade records assignment before 1910.
45The category “Exports to other countries” has been omitted because its inclusion didn’t permit us to
control for partner country characteristics.
43
industrial capacity nor with changes on market characteristics, but simply with a change
in the classification rules followed by the official statistics. For an extended discussion of
this point related with the extensive margins of trade specification, see Appendix B.
In addition, it is important to clarify that we also include the zero values of
countries to which Japan exported in a determined period, so zeros are included in our
database and in the main regressions.
1.2 Main Colonies bilateral trade data with Japan
The main drawback of the official Japanese trade records is that they don’t include
data on Japanese exports towards its main colonies, Korea and Taiwan (although,
thankfully they account for exports towards Kwantung province and Manchuria both
before and after being occupied for Japan). In order to solve this relevant inaccuracy of
the statistical records, and believing that the study of Japanese export patterns towards its
colonies is vital for our research target, we have obtained information on Japanese exports
towards Taiwan from the Returns of the trade of Taiwan for forty years (1896-1935) and
annual Return of trade of Taiwan (Formosa) (1936-1942) both published by the
Department of Finance of the Government of Taiwan. Such returns also include
Taiwanese imports from Japan (or equivalently Japanese exports to Taiwan) at the product
level. The quality of this database is acceptable; however the level of disaggregation is
not the same as the one of the Annual Return of the Foreign Trade of the Empire of Japan
as it includes a lower disaggregation by product, a fact which could underestimate the
extensive margin for Japanese colonies. This is a problem in our data that remains
impossible to fullysolve, but we have taken several steps to remedy the problem. First,
we take the bilateral trade records of Japan with Taiwan in 1915 as a base year because it
is the last year in which the number of goods exported to Taiwan employs the same
disaggregation level as the other Japanese bilateral records, and then we assume that every
3 SITC category follows the same tendency as the respective ones of the Kwantung
Leased Territories.
For the case of Korea, the Annual Returns of the Foreign Trade of the Empire of
Japan in their volumes from 1914-1920 include an Appendix with data on Japanese
exports to Korea disaggregated at the same product level as Japanese exports to the rest
of the countries. For the rest of the period, however there is no alternative available source
in English that breaks down Japanese exports to Korea at the product level. This is
problematic for our research because Korea was probably the most important colony
44
throughout the period. In order to solve this we have obtained from the Long Term
Economic Statistics of Japan (LTES tables J1411__001 to J1411__007) at Hitotsubashi
University (http://www.ier.hit-u.ac.jp/English/databases/ltes.html) a collection of
Japanese exports to Korea disaggregated in this case at the 5 sector level (which include:
I Agricultural Marine and Forest Products, II Mineral Products, III Manufactured
Processed Food, IV Textiles and V Chemical, Metals and Machinery) Sectors I and II
correspond to Primary products according to our skill classification, sectors III and IV to
our low skilled manufactures and sector V to our high skilled. Finally, in order to obtain
the extensive margin of Japanese exports to Korea for 1925-1938 we have used, as in the
case of Taiwan, 1915 as a base year of skill disaggregated margins and assumed that the
number of goods exported inside each sector follows the same tendency as the one of
Kwantung, so as to keep the skill breakdown of total exports to Korea unaltered.
We make a conscious decision on the uncontrolled potential bias introduced by
this in our data base, but we believe this is our best option and an acceptable assumption
for several reasons: first, because Kwantung is the only stable Japanese colony throughout
the period, it is affected by the same political changes as Korea and Taiwan. Kwantung
was at first strongly influenced by the railway construction like Korea and the economic
structures of those three colonies were similar, being its industrialization after the 1930s
resemblant to the one experienced by Korea; second, because Kwantung is across from
Manchuria, the territory receiving the biggest number of Japanese product varieties
records, rising from 339 products received in 1912 to 911 in 1938, by employing it as a
reference we are computing an upper bound on Korea and Taiwan’s extensive margin
growth.
2. GDP domestic demand
Information on GDP and GDP per capita is taken from the last version of the
Maddison project and refers to GDP in 1990 GK dollars or real GDP (see Bolt and Van
Zanden (2014). This database does not provide information for every country to which
Japan exported during the period 1912-1938, but it includes information for future
periods. This information is interpolated back to the interwar years by assuming that they
have followed the same rate of change as total exports of the corresponding countries (in
constant prices) obtained from Federico and Tena-Junguito (2018, 2019). This last
method is far from perfect because exports didn’t grow at the same rhythm as total GDP
but is a method generally accepted for computing upper bounds on GDP growth.
45
Another addition to the literature is the estimation of GDP levels and other
statistics for Japanese colonies of Kwantung Leased Territories and Manchukuo.
Population data has been obtained from Kang (1981) and the corresponding GDP is
calculated as the proportion of Japanese GDP equal to the percentage each population
represents over the Japanese total (the exercise for Manchuria is the same but using China
as a reference because until 1932 it was part of China). The Kwantung GDP is considered
to follow the same rate of growth as the Japanese one (since it is a Japanese colony),
whereas Manchuria’s follows the same tendency as China until 1932.46 Afterwards, its
GDP is assumed to grow at the same rate as Manchurian exports taken from Federico and
Tena-Junguito (2018, 2019).
There still remain a group of territories that are provinces, colonies or autonomous
territories belonging to European Empires for which no data is available but which appear
as independent entities on Japanese statistics and for that reason are treated as such in the
present research. So, we use also marginal data recorded for Africa and the Caribbean
polities. We include their GDP proxied as a proportion of its colonizer’s GDP, according
to total population. Population from those territories is obtained either from the League
of Nations or from censuses obtained on the mainland official statistics (like Annuarie
Statistique de la France, Spanish INE data, Annuario Statistico Italiano, British
Parliamentary papers etc…). This procedure allows us to extend GDP estimations for as
many countries as possible in order to get partners’ demand by their respective economic
size growth (dividing it by population obtained from the League of Nations and Mitchell).
3. Productivity estimation and sources
3.1 GDP per hour worked differentials
Productivity is estimated through the division of GDP by the corresponding data
on total hours worked found mostly in Huberman and Mins (2007) in order to construct
GDP per hour worked. Working hours refer to total hours worked per worker (male and
females) in productive non-agricultural activities. This first source only includes data of
hours worked per employee for 15 advanced countries. Our database is complemented
with information on total working hours for 9 peripheral countries in Europe including
46Kang includes data for 1912 and for the period 1924 to 1941. Population for Manchuria in 1915 has been
obtained by multiplying 1912 data by the rate of change of total Chinese population and for the case of
Kwantung Leased Territories the same has been done but assuming the same rate of change of Japan’s
population (as it was a Japanese colony) in order to differentiate development of Japanese colonies from
other East Asian territories
46
Russia obtained from Maddison (2007) and Maddison (1995). Information on both
databases is not annual, so it has been translated to our benchmark years assuming
constant annual rates of change.
This data is likely to refer to a subset of economic activities, excluding agriculture
that is a sector with complex seasonal working hours. Huberman (2007) and Maddison
(1995) would show that the average number of hours per worker decreased substantially
during the first third of the twentieth century. We know, according to Federico (2005) that
the number of hours and part time work in advanced countries reduced the number of
working hours in agriculture, probably at a slower rate, but adapting to the general trend
of the wide economy in the period. Unfortunately, Federico’s (2005 Table 4.2 p.63) data
on agricultural working hours is too scarce and provisional to be used to introduce a
different correction for agriculture than for the general data on the wider economy. In
order to avoid measurement errors in the construction of the variable we have decided to
deduct agrarian output from total Maddison GDP. Agrarian share over National product
is found in Mitchell (2003) and we subtract the corresponding percentage from
Maddison’s data in 1990 GK dollars. To estimate non-agrarian GDP per hour worked for
those 24 countries and for Japan we multiply hours worked on average by the employment
in the non-agrarian sector which is also obtained from Mitchell (2003).
Finally, all this information is complemented with data on GDP per hour worked
for Taiwan (representative of productivity in Japanese colonies), Indonesia (representing
productivity in Japan’s future dominions), India (which is the leader among other Asian
poor territories outside the Japanese sphere of influence) and bigger South American
countries like Argentina and Brazil. We use total hours per worker for those territories in
1950, from Maddison (1995), multiplied by non-agrarian employment obtained in
Mitchell (2003). We interpolate back in time assuming they follow the same trend as total
employment (excluding agriculture), following the same source.47 All in all, GDP per
hour worked in non-agrarian sectors is a variable that in this database is offered for Japan
and 27 country partners which represent 92% of Japanese exports. In order to capture
Japanese productivity improvements relative to partner countries we divide Japanese non
agrarian GDP per hour worked by the corresponding partner’s one.
3.2 Real wages and productivity differentials by country and sector
47 India and Indonesia total hours in 1950 are assumed to be equal to the ones that appear under the
section “other Asia”.
47
We also use relative real wage growth between Japan and its partners as an
additional variable to capture productivity gaps. The use of this variable generates
controversy among scholars, because when there is no perfect competition, as in Japan
during the interwar years, real wage differentials may not reflect productivity
differentials.
The first strategy consists of building a wage index ratio by dividing Japanese real
wage growth in relation with the corresponding partner along benchmarks (so an increase
of the index means that Japanese wages are rising faster than those of its trade partners).
Data for real wages is obtained from Williamson (1995) which includes information for
30 countries on the five continents including the Japanese colonies of Korea and Taiwan
and future conquests like the Philippine Islands, Indonesia or Siam. Regarding countries
with no information, as is the case of GDP per hour, we include a zero value.
The second strategy is based on the assumption that Japanese comparative
advantage was not necessarily related with higher wages that might reflect higher quality
production, but perhaps sectoral lower unit costs. Following this idea, we estimate an
additional productivity measure to capture, as much possible, the differences in the
evolution of relative skill sectoral productivity in Japan. We have found data on Japanese
manufacturing output in different sectors and their respective employment during the
period. We use the machinery sector to capture the evolution of the high skilled sector
and the cotton textiles sector for low skilled manufactures. Data is offered in LTES (Long
Term Economic Statistics of Japan, tables JPA19__008, JPA19__002 and JPA16__008
respectively) from Hitotsubashi University. We use the employment share of each sector
over total hours worked as a proxy of the number of hours worked in every sector. This
allows us to construct two productivity differential indices for Japan in high skilled and
low skilled manufactures. We use each of them divided by partners’ overall productivity.
Lastly, information regarding population density as a proxy for the difference
between Japan and its commercial partner’s factor endowments has been obtained from
League of Nations (1927-1945) Statistical Yearbooks data.
4. Trade costs
4.1 Transport costs
The most common variable employed in the literature on gravity to measure
transport costs is distance between countries, but it is time and product invariant. This
48
paper makes an effort to overcome the implications of this limited assumption on the
measure of transport costs. A relevant contribution of this paper is the construction of a
database of estimated freight factors (cost of transport per ton/product price per ton)
between Japan and each country partner, for every single product exported by Japan for
each of our corresponding benchmark years. In our strategy geographical distance
between Japanese partners is combined with different freight rate routes and differences
in freight factors by composition of products. freight rates are better proxies for transport
costs than simple geographical distance, as they reflect the variations in number, size,
tonnage, speed and oil consumption on vessels employed in the different routes or the
freight price differentials driving by shipping competition along each route.
In order to do so, we have first delimited the world into three freight regions:
Europe (including Africa), America (North and South) and Asia (including Oceania), the
latter crucial for the study and subsequently divided into three further regions: the East
Asia route, which includes those Asian countries which were closer to Japan and with
which it had strong colonial or other political relationships, the Southeast Asia route,
including countries conquered by Japan and annexed to the Empire during WWII, and the
Other Asia and Pacific route, with very distant Asia and Pacific countries with which
Japan did not have any economic nor political stake.
Those freights also vary between product sectors since we have a different freight
factor for every 3-digit- SITC, making our trade costs variable sensitive to changes in
trade composition by trade partners and in consequence by skill level. We are able to
compare these results only for the general equation without skills with those obtained
using the conventional method of Great Circle Distances (in kilometres) between Capital
Cities (Source: own calculations using Latitude and Longitude data from worldatlas.com
), and in this case the results are very similar, with coefficients presenting almost the same
levels and significance as is shown by Table 3, the coefficient of Colonies and Future
Conquests being even more powerful.
Table 1: Main Regression Employing Distance Instead Of Freights.
(1) VARIABLES EXP
GDP 1.351***
(0.424)
49
RELPRODUCTIVITY 0.479
(0.395)
REL WAGES 0.460
(0.389)
POPDENSITY DIFF 0.544***
(0.179)
GDPCAP ABSDIFF 1.516***
(0.453)
DISTANCE 0.157
(0.280)
EXCHCONTROL -1.081
(0.772)
EXCHRATE -0.468***
(0.124)
TARIFF -0.357**
(0.178)
COLONY 2.858***
(0.983)
OCCAFTER1938 1.006**
(0.391)
DIPLOMATS 0.116
(0.101)
IMPORTS 0.162
(0.139)
Constant -9.224
(5.819)
Observations 674
R-squared 0.636
The procedure for obtaining our freight factor data base is the following one: we
have first delimited the world into three freight regions or transport cost routes, assuming
a common freight per route which will vary between countries according to their distance
with Japan. The main Japanese trade routes in our paper are Europe (including Africa),
America (North and South) and Asia (including Oceania). The latter route is crucial for
the study and subsequently is divided into three further regions like the East Asia route,
which includes those Asian countries which were closer to Japan and with which it had
strong colonial or other political relationships, the Southeast Asia route, including
countries conquered by Japan and annexed to the Empire during WWII, and the other
Asia and Pacific route, with more distant Asia and Pacific countries with which Japan
didn’t have any economic nor political stake. We have 5 routes in total.
The next step was to find at least a representative freight rate (cost of transport per
single ton of a determined product to a specific destination) for each route during the
period 1912-1938. Cost of transport per ton mileage has been adapted to the different mile
distance of Japan’s trade partners for each route. East Asian freight rates were found in
Yasuba (1978) and were based on the freight of coal between Nagasaki and Shanghai for
the years 1910, 1920, 1930 and 1935. From this source we also taken the freight rate of
50
cotton between Japan and Bombay in 1912, which is employed as the basis for computing
the Other Asia and Pacific freight rates. In both cases, the corresponding freights are
translated to our benchmark years by using an index offered also in Yasuba’s charts,
which covers the years 1880-1940. Freights for the Southeast Asia route are found in
Ellinger, B., & Ellinger, H. (1930) and are based on cotton manufactures freights between
Osaka and Batavia in 1929, which are then translated to the rest of the years of interest
using Shimizu, H. (1988), which describes the percentage variations of general freights
between Japan and Indonesia from 1902.
Additionally, for the European and American routes for which more freight
information is available, we have chosen two freights as references: light ( where the
transport cost is small compared to the price) and heavy freights (where the opposite
occurs). Specifically, they are tea on the London-Shanghai route (found at “Freights
1800-1938” http://www.uc3m.es/tradehist_db in Federico& Tena-Junguito (2018)) and
coal on the Wales-Hong Kong route obtained from Isserlis (1938), respectively, for the
case of Europe. For the case of America, the freights are raw cotton and iron and steel
between USA and Japan obtained from Sanderson (1926) and Sanderson (1940). It is
necessary to mention that the different freight rates employed are denominated using
different currencies and units of measure. However, all of them have been translated into
sterling pounds per ton by using market exchange rates from Federico & Tena (2018)
after taking equivalences between different units of measurement.
When we had obtained a complete series of freight rates (in pounds per ton) for
each of our five routes, we translated them to every country belonging to each route. The
procedure followed consists of dividing every route representative freight by the total
distance covered (for example the East Asia freight is divided by distance between
Nagasaki and Shanghai) in order to calculate the route specific cost of transport per ton
and per mile.
TABLE 2. Japan Freight rates in different routes (1912-1938)
EAST ASIA SOUTHEAST ASIA
SOUTH ASIA
EUROPE AND AFRICA LIGHT
EUROPE AND AFRICA HEAVY
AMERICAS LIGHT
AMERICAS HEAVY
1912 0.000135 0.000099 0.000217 0.000199 0.000169 0.000352 0.000176
51
1915 0.000190 0.000180 0.000213 0.000285 0.000208 0.000455 0.000506
1925 0.000188 0.000399 0.000615 0.000096 0.000167 0.000523 0.000299
1929 0.000176 0.000399 0.000607 0.000127 0.000175 0.000429 0.000245
1932 0.000133 0.000319 0.000504 0.000202 0.000109 0.000663 0.000379
1938 0.000328 0.000528 0.000488 0.000162 0.000113 0.000438 0.000166
Data: in Pounds per Ton per Mile
Source see text.
The next step was to multiply each of the route specific freight rates above by the
distance between Japan and the different countries inside every route, in order to build
the freight rates of Japanese exports (pounds per ton) to each country. These freight rates
are then divided by the international price in pounds per ton of each selected commodity
(found in Federico & Tena-Junguito (2018, 2019). By doing this we obtain freight factors
(cost of transport as a percentage of products’ price for 5 goods (coal, cotton
manufactures, raw cotton, tea and iron and steel).
Finally, to vary transport costs across different sectors we translate those 5
benchmark freights to the rest of products. We apply freight factor equivalences between
products obtained in Moneta, C. (1959), which employs freight factor equivalences for
almost 80 products imported by Germany in 1951. The procedure consists of multiplying
the freight factor of each route’s reference product (remember, five routes, five different
reference products) by the proportion that each corresponding good represents over the
reference one. As an example, if the freight factor of meat is 5.3 times bigger than tea’s
freight factor (one of our references), then we multiply tea’s freight factor by 5.3 in order
to obtain the freight factor for meat exports. This process is repeated for each product. As
a rule we assume that products sharing a 3-digit SITC classification have the same freight
factors. In addition we also assume that freight factor equivalences did not change too
much between our period (1912-1938) and 1951.48 Graph 1 shows the evolution of the
freight factors created on this article by means of an index number (1912=100) capturing
changes in the average freight factors for each trade route. Those averages have been
weighted according to the proportion that each corresponding product possesses in total
48 That’s a strong assumption but this is the best possible option for translating our freight factors to every
single product exported by Japan.
52
Japanese exports. The graph shows how freights fluctuate considerably, reflecting
technological, market and trade composition factors, as opposed to distance, which is
fixed. Furthermore, it is remarkable to observe that during the 1930s, the period in which
Japanese exports towards East and Southeast Asia experienced their sharpest raise, freight
factors to both routes increased.
Figure 1. Freight Factor (trade-weighted by route) 1912-1938 (1912=100)
Source see text
4.2 Tariffs, diplomats and exchange controls
Average levels of tariff protection come from Blattman et al. (2003) and represent
total import duties collected by each country as a percentage of total imports.49 This
database has been extended following the same procedure using data obtained from the
British Parliamentary Papers and the Annuarie Statistique de la France for their
respective colonies all over the world (including Africa and the Caribbean). For other
countries, we assume that levels of protection are the same as those of their respective
Metropolis.50 All in all, we have data on tariffs for every country to which Japan exported
during the interwar years.
Finally, we have obtained a subset of other variables representing trade costs such
as exchange rates by country with the Japanese Yen (relative exchange rates from
49Tariffs between Japan and its colonies are assumed to be zero. 50 An example of this could be the Canary Island on Spain. For an extended discussion on tariffs sources
and methods see section 4.2 Appendix A.
0
0,5
1
1,5
2
2,5
3
1912 1915 1925 1929 1932 1938
East Asia SouthEast Asia South Asia Europe and Africa Americas
53
Federico and Tena-Junguito (2018, 2019). This source gives nominal exchange rates
between local currencies in different territories (among which Japan is included) and the
dollar, expressed in local currency/dollar. The new rate is expressed in local currency per
Yen and obtained through the division of partners’ currency nominal exchange rate
(foreign currency per dollar) by Japanese nominal exchange rate (Yen per dollar) and has
been created for 44 partners. In other words, an increment of this variable reflects a
relative nominal appreciation of the Yen. This assumes that nominal devaluation with the
Yen affects trade independently of the behaviour of their respective partners domestic
price. Unfortunately, there is a scarcity data on domestic prices for the cross-section of
countries, which impedes the inclusion of bilateral effective real exchange rates in our
gravity equations. Furthermore, we have also obtained data on exchange rate controls
obtained from various sources including Bethell (1994), Eichengreen (1996), Meisel
(1990) and Reinhart & Rogoff (2002).
Finally, we collect information on the number of Japanese diplomats operating in
each country, which has been obtained from Almanach de Gotha (various years), which
includes a section with the names and country of origin of every diplomat operating in
each country. We have meticulously counted the number of Japanese diplomats that
worked in each country and the number of foreign diplomats that worked in Japan.
Table 3 Number of Japanese Diplomats Abroad
1912 1915 1925 1929 1932 1938
AFGHANISTAN
3
KENYA
1
FINLAND
1 1 2
MORROCCO
1
LEBANON
1
IRAN
1 4 5
ITALIAN AFRICA
1
GERMANY 15 RUPTURE 23 17 13 15
LATIVA
2
USA 21 22 33 27 23 22
HAWAII 1 1 1 1 1 1
PHILIPPINES 1 1 1 1 1 2
ARGENTINE 1 1 4 4 5 7
AUSTRIA-HUNGARY 8 RUPTURE 7 5 3 3
54
BELGIUM 6 6 12 8 7 6
BOLIVIA
1 1 1 1
BRAZIL 5 5 11 10 10 11
CHILE 4 4 6 5 5 5
CHINA 40 38 67 68 60 20
COLOMBIA
1
3
CUBA
1
1 1
DENMARK 1 1 1 2 2 1
EGYPT
3 2 2 5
SPAIN 4 3 7 4 6 3
FRANCE 13 10 25 22 18 13
FRENCH INDOCHINA 1 1 2 3 2 2
GREAT BRITAIN 16 14 32 21 17 19
INDIA 2 2 3 4 5 4
STRAITS
SETTLEMENTS/MALASYA 1 1 2 1 1 2
CANADA 2 2 3 2 6 8
AUSTRALIA 3 2 6 1 1 2
NEW ZEALAND 1
2
SOUTH AFRICA 1
1 1 1 1
HONG KONG 1 1 2 1 1 2
SRI LANKA 1 1 1 1
1
GREECE
2 2 2
HUNGARY
1
ITALY 13 14 16 14 9 9
LATVIA
3
1 2
LUXEMBOURG
1
1
MANCHUKUO
1
15
MEXICO 3 6 6 7 8 8
NORWAY 1 1 1 1 1 1
PANAMA
1 1
1
NETHERLANDS 5 5 7 5 4 5
INDONESIA 1 1 2 2 1 1
PERU 1 4 6 4 4 6
POLAND
4 6 4 6
PORTUGAL
4 3 1 4
ROUMANIE
5 4 2 2
EL SALVADOR
1
RUSSIA 14 18 1 23 18 16
SIAM 3
3 4 3 2
SWEDEN 4 5 10 6 5 2
SWITZERLAND 1 1 9 4 3 4
TURKEY
1 9 6 6
URUGUAY
1 1
CZECHOSLOVAKIA
2 5 2 2
TOTAL 195 171 344 314 266 256
Sources: Japan’s Diplomats abroad comes from Almanach de Gotha (various years)
55
Table 4: Number of foreign diplomats in Japan
1912 1915 1925 1929 1932 1938
AFGHANISTAN
2
EGYPT
1 1
CANADA
1 5 4
HONDURAS
1 1 1
IRAN
1
LUXEMBOURG
1 1 1
MANCHUKUO
11
NICARAGUA
1
DOMINICAN REPUBLIC
1
EL SALVADOR
1
1
GERMANY 15 RUPTURE 8 11 10 12
USA* 20 19 32 18 26 35
ARGENTINA 3 3 2 3 5 2
AUSTRIA-HUNGARY 5 RUPTURE 1
1 1
BELGIUM 7 6 7 4 2 4
BOLIVIA
2 6 4 2 3
BRAZIL 9 6 7 4 6 5
CHILE 8 4 6 3 4 5
CHINA 20 16 19 12 14 30
COLOMBIA
2 1 1 2
CUBA
4 1 3 2
DENMARK 3 4 8 5 6 1
ECUADOR
1 1
SPAIN 6 5 6 4 7 1
FINLAND
2 2 3 2
FRANCE* (1) 10 10 10 12 10 14
UK 24 22 21 19 20 19
GREECE
1 3
1
ITALY 5 6 6 7 7 8
MEXICO 6 6 3 5 3 4
NORWAY 6 5 8 2 4 2
PANAMA
1 1 1 1
PARAGUAY
1
1 1
NETHERLANDS 7 9 8 5 8 5
PERU 2 2 5 6 3 4
POLAND
3 6 4 3
PORTUGAL 3 5 8 2 2 2
PERSIA
2
ROUMANIA
1 3 3 4
RUSSIA* (3) 21 19 1 12 14 14
SIAM 2 2 3 3 4 4
SWEDEN 4 2 10 5 3 3
SWITZERLAND 2 2 2 2 1 2
CZECHOSLOVAKIA
3 3 2 3
TURKEY
1 1 2 5
URUGUAY
1
56
VENEZUELA
1 2
1 1
TOTAL 188 157 211 171 195 228
Sources: Foreign diplomats in Japan comes from Almanach de Gotha (various years)
4.3 International demand by Imports
Lastly, we believe that the general international protectionist backlash prevalent
in Europe,the USA, and other relevant markets for Japan, such as India, during the 1930s
influenced Japan’s commercial focus on its Empire and regional markets. In order to
capture the global trade diversion generated by the Great Depression, we have included
in our database a variable capturing the contraction of international trade over domestic
demand represented by the GDP. It is represented by total imports for every Japanese
country partner in each of our benchmark years. Information is obtained from Federico
and Tena-Junguito (2018,2019) and refers to total imports in current dollars. This source
includes information for almost every Japanese trade partner. Exceptions are Hong Kong,
Kwantung and Manchuria (1912-1929, our source contains information for 1932-1938),
which were part of Japan’s future dominions and colonial possessions, respectively. In
order to account for these omissions, we have obtained Hong Kong imports from the
British Parliamentary Papers (Statistical Abstract for the British Empire 1924-1938). The
rest of the years have been extrapolated assuming Hong Kong exports to follow the same
path as British Malaysia as both were British Colonies and were famous entrepots (British
Malaysia includes the Singapore entrepot). To calculate Kwantung imports, we assume
them to be equal to the proportion of Japanese imports based on the percentage of the
population of Kwantung represented over the Japanese population and assuming parallel
trends. Finally, Manchurian imports have been interpolated back from 1932 to the
previous benchmarks by assuming imports to evolve at the same rate as its population.
4. 4 Japanese colonies and future conquest territories’ main features
Among the variables chosen to be part of the model there are two that are related
mainly with political factors that imply severe changes in trade costs related with changes
in the connection of different territories within the Japanese Empire over the period. As
has been explained in the literature review section in the paper, this two forms: the first
one consisted of the Japanese creation of formal colonies in the conventional way and the
second one consisted of an economic penetration that anticipated future conquests. We
57
believe that both kinds of imperialism allowed Japan to reduce trade costs with its
neighbours and we build two separate dummies for representing them. The first one
captures Japanese trade bias towards its colonies, whereas the second tries to measure the
extent to which Japan’s interest towards its future conquered territories implied bigger
exported quantities than what is suggested by the rest of the model’s variables. The
following tables summarizes which countries were part of Japan’s Colonies and future
dominions and their main characteristics.
58
MAP.1: JAPANESE CONQUESTS DURING IMPERIAL PERIOD (1880-1945)
59
Table 5: Japanese Colonies Status and Main Features.
Japanese Colonies (Formal Empire)
Annexation Date
Period As Colony
Distance (Nautical
Miles) With Japan
Pop 000s (1912)
Pop 000s (1938)
GDP Mill. 1990
Dollars (1912)
GDP Mill. 1990
Dollars (1938)
Tariffs
(1912)
Tariffs (1938)
Share Over
Japan's Exports (1912)
Share Over
Japan's Exports (1938)
Main Product Traded (1912)
Main Product Traded (1938)
Taiwan 1895 1912-1938
920 3,411 5,552 2,297 7,252 0% 0% 7% 8% Cotton Cloth
Sulphate Of Ammonium
Korea 1905 (Protectorate
) 1910 (Colony)
1912-1938
449 10,422 15,381 8,789 24,895 0% 0% 8% 23% Gray Shirtings
Iron Manufactures
And Gray Shirtings
Kwantung Leased Territories
1905 1912-1938
920 596 1,750 218 544 0% 0% 4% 14% Gray Shirtings
Metals And Machinery
Manchuria 1931 1932-1938
768 30,857 40,354 11,288 42,307 0% 0% 0% 8% No Trade Machinery And Wheat Flour
China 1937 1938 735 432,375 513,336 158,176 288,653
3% 27% 19% 8% Cotton Yarn
Spinning Machines And Wheat Flour
Notes: Tariffs represent tariffs with Japan, except in the case of China in which they represent average tariff
Table 6: Japanese Future Conquests And Main Features
60
Future conquests (informal empire)
Annexation date
Periods as informal empire
Distance (nautical
miles) with japan
Pop 000s
(1912)
Pop 000s
(1938)
Gdp millions
1990 dollars (1912)
Gdp millions
1990 dollars (1938)
Tariffs (1912)
Tarifs (193)
Share over
japan's exports (1912)
Share over
japan's exports (1938)
Main product traded (1912)
Main product traded (1938)
Manchuria 1931 1912-1929 768.4 30,857 40,354 11,288 42,307 0% 0% 0% 8% No trade Machinery and wheat
flour
China 1937 1912-1932 734.9 432,375
513,336
158,176 288,653 3% 27% 19% 8% Cotton yarn Spinning machines
and wheat flour
Hong kong 1942 1912-1938 1,341 450 1,479 164 300 0% 1% 4.7% 0.4% Coal and cotton
Coal
British malaya
1942 1912-1938 2,777 3,025 5,207 2,486 7,089 3% 17% 1.5% 0.1% Coal Railways wagons
British borneo
1942 1912-1938 2,496 208 301 171 303 11% 20% 0.0% 0.0% No trade Iron manufacture
French indochina
1940 1912-1938 1,760 16,990 23,164 433 838 16% 8% 0.1% 0.1% Coal Asphalt and raw silk
Indonesia 1942 1912-1938 3,264 51,105 71,484 42,818 80,044 5% 10% 0.7% 2.6% Coal and matches
Shirts and other
textiles
Philippines 1942 1912-1938 1,432 9,206 15,934 8,390 22,948 12% 12% 0.9% 0.8% Coal Cotton under-shirts
Siam 1941 1912-1938 2,268 8,559 14,980 3,012 2,380 7% 28% 0.2% 1.0% Steam vessels
Steam vessels
Burma 1942 1912-1938 2,359 12,220 16,145 7,348 11,942 13% 18% 0.0% 0.4% No trade Cotton tissues
Gilbert and ellice
1944 1912-1938 3,422 31 34 162 451 17% 20% 0.0% 0.0% No trade Jeans
New guinea 1942 1912-1938 2,484 380 670 1,979 4,186 16% 16% 0.0% 0.0% No trade Cotton manufacture
Notes: Tariffs represent tariffs with Japan, except in the case of China in which they represent average tariff
61
APPENDIX B
Number of products and the Extensive Margins of Trade
One of the biggest concerns regarding the exploitation of historical margins of trade
records is related with the treatment of statistical sources. For example, the easiest way to
compute trade margins is to assume that the extensive margin is represented by the total number
of different goods exported by a country. This simple procedure was followed by Huberman &
et al (2017) and we have replicated it as shown in Table 9 of the paper.
Nevertheless, the granular analysis of Japanese margins of trade might be biased
because Japanese official records experienced an important increase in the number of products
recorded. This phenomenon is wide spread across international trade records and it is related
mainly with the increased diversification of tariffs on imported manufactures, the general
improvements in trade bureaucracy and the standardization in the commodity classification of
international records promoted by the League of Nations Brussels convention of 1913 and
during the interwar years (see chap. 4 of Allen and Ely (1953)).
In this sense, increases in product coverage could overestimate our extensive margin
of trade during the 1930s. This means that 1938 accounts for different product varieties of the
same good, artificially incrementing the number of goods exported by Japan.51 If the extensive
margin collects the number of new products exported by Japan, it will be much higher in 1938
than in 1932, but this could merely reflect statistical modification rather than a structural change
in Japanese exports.
Figure 5 provides partial confirmation of this phenomenon. The number of product
varieties in the same sector exported to different countries increased in all cases around 100%
between 1932 and 1938, thus artificially raising the number of goods exported by Japan.
51 A casical example: cotton fabrics may be subdivided into unbleached, bleached,printed, dye . It is one
commodity or four? Anothe example Egypt in its export estatics list nine varieties of raw cotton how many
commodities are one or nine? (see cp. 4 Allen and Ely (1953)
62
Figure 2: Japanese average number of products included in 3-digit SITC sector by region
Source: Annual Return of trade for the Empire of Japan (various years).
The database constructed includes the value of Japanese exports (in current yen) of each
product exported to each country. Each different product exported has been categorised using
the 2nd Revision of the Standard International Trade Classification (SITC) at a 5-level digit
disaggregation in order to achieve a proper and precise product classification. This
categorisation also permits us to distinguish new export products from those exports that are
simply different varieties of the same product. For that purpose, products sharing the same 3-
digit classification compose a single product category or sector used in this paper. This is similar
to the strategy employed by (Meisner & Tang 2018). Through this method we observe that
Japanese industrialization and imperial mechanisms, along with reductions in trade costs,
permitted Japan to increase exports from 117 to 164 sectors, as can be observed in figure 3.
Figure 3: Japanese number of sectors by region
Source: Annual Return of trade for the Empire of Japan (various years)
63
Table 6: Japanese Export Determinants at the Margins of Trade
(Extensive Margin as Total Number of Products)
Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
(1) (2) (3) (4) (5) (6) VARIABLES Total
Extensive
Total
Intensive
Extensive
HS
Extensive
LS
Intensive
HS
Intensive
LS
GDP 0.325*** 1.428*** 0.245** 0.164 0.534* 0.958***
(0.120) (0.341) (0.110) (0.116) (0.292) (0.322)
RELPRODUCTIVITY -0.0880 0.146
(0.290) (0.320)
REL PRODUCTIVITY
HS
0.189 0.623***
(0.141) (0.183)
REL PRODUCTIVITY
LS
-0.239 0.612***
(0.324) (0.173)
RELWAGES 0.180 0.697**
(0.215) (0.285)
RELWAGES HS -0.287 0.381
(0.199) (0.354)
RELWAGES LS 0.474** 0.459
(0.202) (0.374)
POPDENSITYDIFF 0.0325 0.346* 0.0262 0.0829 0.240* 0.170
(0.0666) (0.206) (0.0805) (0.0824) (0.133) (0.152)
GDPCAPABSDIFF 0.0668 1.016** 0.153 0.0975 1.139** 0.0419
(0.260) (0.397) (0.262) (0.252) (0.474) (0.329)
FREIGHTS -0.287** 0.0842
(0.140) (0.308)
FF HIGH SKILL -0.458*** 0.0980
(0.121) (0.192)
FF LOW SKILL -0.414*** -0.382
(0.141) (0.270)
EXCHCONTROL -0.295 -1.173 -0.623** -0.413* -1.258* -0.726
(0.239) (0.875) (0.270) (0.233) (0.660) (0.530)
EXCHRATE -0.168*** -0.316*** -0.176*** -0.162*** -0.337* -0.340***
(0.0531) (0.105) (0.0606) (0.0493) (0.175) (0.0826)
TARIFF -0.177* -0.528***
(0.0963) (0.181)
TARIFFS HS 0.0171 -0.0319
(0.0746) (0.0861)
TARIFFS LS 0.00926 -0.526***
(0.0786) (0.120)
COLONY 0.835* 1.782** 0.762** 0.458 4.091*** 3.270***
(0.442) (0.875) (0.342) (0.433) (0.436) (0.502)
FUTURE CONQUESTS 0.855*** 0.533* 0.571 0.541 1.438*** 3.413***
(0.246) (0.285) (0.397) (0.395) (0.397) (0.660)
DIPLOMATS 0.141* 0.0177 0.173** 0.153** -0.168 -0.0727
(0.0737) (0.131) (0.0741) (0.0758) (0.154) (0.235)
IMPORTS 0.0291 0.0999 0.0175 0.0293 0.530*** 0.167***
(0.0402) (0.0757) (0.0378) (0.0352) (0.181) (0.0396)
Constant 0.506 -10.21** -0.466 -0.833 -10.87*** -5.229*
(1.712) (4.316) (1.535) (1.515) (2.726) (3.053)
Observations 674 674 676 676 676 676
R-squared 0.596 0.639 0.590 0.521 0.713 0.636
64
APPENDIX C
Manufacturing products by skill intensity and the dynamics of textiles
1. General division of total exports by skill intensity
We have adapted the 3 digit SITC categorization to a contemporaneous group of 16
manufacturing sectors following the British Board of Trade (1905) classification of British
manufacturing exports. We have ranked these according to skill intensity following Tena-
Junguito (2010). Table 1 shows these manufacture sectors ranked from high to low median
earning workers at the turn of the century (High skilled (skill intensity between 11 and 13,
ranked 1-6), Medium skilled (skill intensity between 9 and 10, which are ranked from 7 to 9)
and Low skilled manufactures (between 5 and 8 skill intensity, ranked from 10 to 16). This
strategy, suggested by Num and Trefler (2009), consists of differentiating high and low cut-offs
inside the ranking to categorise in a more effective way high skills and low skill sectors. This
strategy is hopefully less vulnerable to technological change during the first third of the
twentieth century.
We have extended this strategy to primary products as a residual (assuming less skill
intensity than in manufactures). Thus, we work with a simple division between high skill sectors
that were more influenced by the second industrial revolution technologies and those that were
less influenced. We recognize that, on one hand, this division represents a very rudimentary of
capturing skill improvements inside sectors, especially productivity increases in Japanese
textile industries in the 1920s and 1930s (See Wolcott (1994)), but on the other hand, this
strategy captures better Japan’s manufacturing dual aggregate sectoral advance in the second
industrial revolution technologies as an indicator of industrial modernization in the inter-war
years, which is the main objective of the paper.
65
Table 8 Skill intensity ranking
Ranking Description skill intensity
1 Ships 13,01
2 Machinery hardware&c 12,65
3 Paper Manufactures 11,65
4 Silk thrown 11,58
5 Iron Steel Manufactures 11,29
6 Leather and Manufactures 11,00 high cut-off
7 Copper lingots, Cakes, Slabs 10,01
8 Alkali Chemical products 9,64
9 Apparel 9,27
10 Woollen & Worsted Manufactures 7,9 low cut off
11 Linen Manufactures 7,8
12 Cotton Manufactures 7,74
13 Jute Canvas and Sacking 7,04
14 Woollen yarns(stuffs all wool) 6,2
15 Linen Yarn 5,9
16 Cotton yarns undyed 5,8
Sources: See Tena-Junguito (2010)
2. Dynamics of Japanese high- and low-end textiles exports
In order to ascertain how the destination of Japanese exports varied across time, and
which dynamics they followed according to economic and political circumstances, it is
pertinent to exploit the possibilities offered by the granular database employed in the paper. For
example, it has permitted us to further divide Japanese textile exports between high-end or
highly sophisticated textiles and low-end ones. The first category includes apparel or finished
clothing, the skill intensity of which is higher than 8, whereas the second category includes
piece goods, yarns and other unfinished textiles with a skill intensity lower than 8.
66
Figure 4: Share of low- and high-end textiles over total Japanese textile exports.
Source: Annual Return of trade for the Empire of Japan (various years).
We note that most textile exports were low-end; that is, cotton yarns, threads and other
kinds of piece goods represented around 65-70% of the total, in contrast to clothing and apparel,
which represented between 25% and 30% of textile exports.
Figure 5: Distribution of low- and high-end textiles by regions (1912-1938)
Source: Annual Return of trade for the Empire of Japan (various years).
71% 66% 72%64% 68% 70%
29% 34% 28%36% 32% 30%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1912 1915 1925 1929 1932 1938
%LOW END %HIGHEND
45%29% 30%
23% 28%
47%
34%
23%
40%
29%25%
11%
21%
47%30%
48% 46% 42%
0
50000000
100000000
150000000
200000000
250000000
300000000
350000000
0%
20%
40%
60%
80%
100%
1912 1915 1925 1929 1932 1938
REST OF WORLD HIGHEND INFORMAL HIGHEND
COLONIAL HIGHEND Total Real High End Textiles
20% 19% 26% 28% 24%
45%
60% 59% 47% 43%
30%
12%
20% 22% 28% 28%47% 43%
0
100000000
200000000
300000000
400000000
500000000
600000000
700000000
800000000
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1912 1915 1925 1929 1932 1938
COLONIAL LOWEND INFORMAL LOWEND
REST OF WORLD LOWEND Total Real Low End Exports
67
Furthermore, it is interesting to observe the market behaviour of both kinds of exports.
In both cases we can appreciate a transition from a focus on controlled territories to a
distribution outside the Japanese sphere of influence, which ends after 1932, the year in which
Great Britain and other powers diverted trade to their Empires and in which Japan, as a
consequence of this and of its invasion of Manchuria, decided to launch an Empire-based
development strategy.
This transition is more evident in the case of textiles, located on the bottom part of the
skill intensity distribution. Beginning in 1912, it is evident that they are mainly exported
towards countries composing the Japanese informal Empire in Southeast Asia (60%).
According to the dynamics followed by other sectors in the Japanese economy, low-end textiles
were first directed towards Japanese colonies which were politically and economically
controlled by Japan. Once they achieved a certain degree of competitiveness, they were
exported towards countries in the informal empire. After that, we can clearly appreciate a
diversification of low-end textile exports towards the rest of the world, suggesting that Japan
had reached a certain competitive threshold both regionally and internationally. This phase
started in 1929, coinciding with the period in which Japanese low-end textile exports
experienced their largest increase. The transition of high-end textiles is less gradual but also
appreciable and in both cases the biggest increase in exports coincides with the opening towards
new markets. Furthermore, it seems that high-end textiles were introduced sooner to
international markets than low-end ones, suggesting that they were more responsive to market
factors.
All in all, according to the literature and the data we have presented, both low- and high-
end textile exports followed a similar tendency, in which they seem to have followed productive
and international market characteristics, at least to a much higher degree than high skilled
manufactures, which were extremely concentrated inside the Empire. Both kinds of exports
experienced a strong Imperial bias after 1932, even though their final distribution was more
evenly shared across regions than that of high skilled industrial exports.
This intuition can be verified following our gravity model for analysing the determinants
of both high- and low-end textile exports. The variables employed have been the same as the
ones used in the main paper since we haven’t been able to obtain reliable variables presenting
a higher degree of disaggregation. For example, productivity reflects GDP per hour worked in
the overall textile sector related to total GDP per worker in partner countries.
68
Table 7: Determinants of high- and low-end textile exports
(1) (2)
VARIABLES High-End Textiles Low-End Textiles
GDP 0.909*** 0.588*** (0.220) (0.213) REL PRODUCTIVITY
TEXTILES 1.395*** 1.246***
(0.205) (0.167) RELWAGES TEXTILES 0.939*** 0.503 (0.304) (0.330) POPDENSITYDIFF 0.530** 0.554*** (0.230) (0.127) GDPCAPABSDIFF 0.395 0.335 (0.462) (0.380) FFLOWSKILL -0.378* -0.284* (0.210) (0.171) EXCHCONTROL -0.626 -0.468 (0.401) (0.414) EXCHRATE -0.544*** -0.567*** (0.0981) (0.100) TARIFF LS -0.324*** -0.273*** (0.109) (0.0959) COLONY 2.955*** 3.004*** (0.424) (0.364) FUTURE CONQUESTS 2.448*** 2.461*** (0.490) (0.437) DIPLOMATS 0.107 0.303** (0.0843) (0.125) IMPORTS 0.182*** 0.140*** (0.0492) (0.0313)
Constant -4.496 0.488
(2.975) (2.352)
Observations 676 676
R-squared 0.791 0.779
The results show that productivity played a role in determining these kinds of exports,
which is larger than the one it played in other sectors. This suggests that Japanese comparative
advantage played a relevant role in the expansion of textile exports and its strength was similar
at both ends of the skill classification. Contrary to high skill manufactures, textiles were driven
by productivity with the difference that a relevant share of their exports were directed towards
colonies and future conquered territories, although both kinds of textiles were more biased
towards the first than to the second. Interestingly, we can also appreciate that high-end textiles
were slightly more responsive to changes in comparative advantage variables and less sensitive
to imperial controls.
69
Both low- and high-end textiles were first introduced in the colonies, then in the
Japanese future conquests territories, and once they became internationally competitive, they
were distributed all over the world. What’s more, they demonstrated significant expansion in
the world market in spite of the backlash bias towards its colonies experienced after the reaction
of the British Commonwealth trade bloc to the Manchuria invasion that partially excluded
Japanese textiles from India and other markets. In addition, the comparison with second
industrial revolution exports and the fact that productivity played a bigger role in determining
textile exports, along with the dynamics presented above, suggest that textile exports responded
more to productivity determined comparative advantage than high skilled manufactures, which
were more strongly affected by Japanese imperial aims.
70
APPENDIX D
The Great Depression, Imperial policies and manufactured exports
1. The Great Depression, Ottawa agreement and Japanese Empire
The reasons behind Japan’s considerable commercial concentration in its colonies and
neighbouring countries inside the region are varied and not fully reviewed in the paper. Many
authors view the Japanese invasion of Manchuria in 1931 as a turning point in the sense that
after this great victory, the military gained important political power that permitted them to
continue lobbying the government for a continued pursuit of Imperial expansion. Of course,
this historical event was of relevance to the subsequent commercial, economic and political
decisions adopted by the Japanese government during the 1930s, as were the needs of the
Japanese Zaibatsu for gaining access to bigger markets and achieving economies of scale in a
context of a protected market with monopolistic privileges. Nevertheless, there are other factors
that might have also facilitated Japan’s strong focus on its empire as a source of exports. In that
sense, the international landscape might have played a significant role.
First of all, after the 1929 crash, GDPs declined rapidly and the collapse of international
demand was reinforced by a generalized increase of beggar my neighbour policies with
increased tariffs, competitive devaluations, and the creation of new trade blocs. Differences in
international and domestic demand declines are captured in our regressions. In that sense, the
immediate Japanese response was to focus on exerting its comparative advantage on closer
markets, especially in India, which was by far the largest accessible market for Japan apart from
its colonies at that time. This can be appreciated in Figure 3 of the paper (Other Asia and Pacific
share raise from 18% to 24% after 1929) and also in Figure 6 in the present appendix in which
we can see how the vast majority of manufacturing exports towards this region are concentrated
on India.
71
Figure 6: Japanese Manufacturing exports by regions
Source: Annual Return of trade for the Empire of Japan (various years).
Regarding the kinds of products and the competition faced by Japan in its territory, it is
remarkable how Japan concentrated its manufacturing exports to India on textiles as shown in
Figure 6:
Figure 7: Distribution of Japanese exports to India 1929-1938
Source: Annual Return of trade for the Empire of Japan (various years).
As show Table 1 of the text, Japan enjoyed comparative advantage in the production of
textiles. This is also confirmed by Robertson (1990) and Wolcott (1994), who mention that
during the interwar years, especially the 1930s, Japan became the world’s main exporter of
cotton textiles. According to the former author, the main competitor of Japan on the Indian
market was Great Britain, which also enjoyed political privileged access to the territory.
78% 77%61%
15% 13%
20%
7% 10%19%
0%
20%
40%
60%
80%
100%
1929 1932 1938
India Australia Others
18% 20% 22%
79% 76% 75%
3% 4% 3%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1929 1932 1938
High Skills Textiles Others
72
Competition grew in intensity during the 1920s, but Great Britain didn’t give much importance
to the growth of Japanese cotton textile exports to India since they attributed them to the 1924
Yen devaluation. Indeed, time proved the British correct because, after the Japanese financial
crisis in 1927 and subsequent increase in the price of Yen, Japanese exports to India decreased.
Nevertheless, Japanese cotton textile production enjoyed other advantages over the
British than lower labour costs or exchange rate depreciation. The large scale of Japanese textile
producers, the high level of integration between firms, the tendency to concentrate on a reduced
set of product varieties and the closer links between producers, suppliers and merchants,
provided a cost advantage that the British producers couldn’t match. This advantage also
allowed the Japanese to surpass the British in cotton textile exports not just to India, but also in
other markets traditionally dominated by the British, like China or East Africa. All in all, it
seems that the commercial performance of the Japanese outside its Empire was dominated by
textile exports, in which it was the leading exporter, and this performance was possible thanks
to factors awarding the sector comparative advantage.
The largest increase of exports to India during the Great Depression might be also
related with the Yen’s devaluation after 1931, which was larger than that of the sterling pound.
Between 1929 and 1932, Japanese exports towards its Empire experienced a reduction due to
the Chinese boycott in retaliation of Japanese military campaigns over Manchuria in 1931. This
fact could have also forced Japan to diversify its exports towards territories like Australia,
Africa or Ceylon. In each of these territories, Japanese cotton textiles enjoyed a comparative
advantage over British ones.
For this reason, the British response to Japanese competition in cotton textiles in India
and other colonial territories came in the form of imposing higher tariffs and quotas after 1932
and forcing them to negotiate to raise their prices. This would have provoked the retreat of
Japanese exports from British dominions appreciable in previous graphs, and may explain the
strong Japanese focus on its empire which is observed in 1938 exports. Those exports were, as
shown in the paper, not dominated by comparative advantage, but rather by other political
motives and focused on high skilled manufactures.
73
2. Great Depression colonial bias on high skilled manufacturing exports
We deal here with a potential bias of the effect of the colonial increase on total Japanese
exports despite their low initial levels. It is pertinent to offer evidence that proves that the
colonial effect is much bigger for high skilled exports.
In order to do, we have calculated increases in total, manufacturing and high skilled
exports for the period in which they rose the most (1932-1938) and have computed the colonial
share of these increases. The results clearly show a colonial bias, as colonies represent 86% of
the increase in total exports. This proportion is slightly reduced when we examine the increase
in total manufactures (84%). But the evidence is conclusive, showing that Japanese colonies
were by far the main destination market of high skill manufactured exports after 1932,
representing 96% of the total increase. In other words, figure 8 shows the colonial bias of
Japanese total and manufacturing exports and that this bias was much higher for high skilled
manufactures.
Figure 8: Japanese exports increase 1932-38 (%) by groups of products and region.
Source: Annual Return of trade for the Empire of Japan (various years).
86% 84%96%
14% 16%4%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Total Exports Manufactures High Skills
Colonies Rest of the World
74
APPENDIX E
The endogeneity of exports on productivity and transaction costs.
Another important concern related with this article and with papers in general dealing
with export determinants and productivity is precisely the possible endogeneity between the
key variables. For example, productivity increases may be affected by export expansion. The
mechanisms are related with the fact that exogenous changes in trade costs might also generate
improvements in productivity through economies of scale (the market expands and unit costs
reduce) or experience (increasing production might lead to learning by doing and reduced
costs). For that reason, the first test we are going to run in this section is to check whether
increases of exports generated improvements in productivity in Japan.
Table 9: Effect of export increase on productivity
(1) (2) (3) (4) (5)
VARIABLES Rel productivity Rel productivity Rel productivity Rel productivity IV
logrelativeopen -0.0721
(0.0795)
logexports 0.0418
(0.0261)
logexportsHS 0.0572**
(0.0257)
logexportsLS 0.0949***
(0.0344)
ivexports -0.0251
(0.0429)
Constant -0.381*** -0.877** -0.367 -1.293*** 0.177
(0.0482) (0.368) (0.315) (0.443) (0.673)
Observations 123 140 131 133 149
R-squared 0.012 0.034 0.068 0.135 0.003
In order to do so we follow an approach similar to that of Huberman et al. (2017): first,
we check whether improvements in openness (increases of Japanese exports as a percentage of
GDP) generated increases of Japanese productivity (measured as non-agrarian GDP per hour
worked). The same has been done using nominal exports or high and low skill manufacturing
exports. Finally, an instrument has been created which is called ivexports and is constructed by
the predicted value of exports regressed on the usual variables with the exception of
productivity measures.
75
IVEXPORTS=𝛽0 + 𝛽1𝑙𝑛𝐺𝐷𝑃𝑖𝑡 + 𝛽2𝐶𝐴𝑖𝑡 + 𝛽3𝑇𝐶𝑖𝑡 + 𝛽4𝐶𝑂𝐿𝑂𝑁𝑌𝑖𝑡 +
𝛽5𝐹𝑈𝑇𝑈𝑅𝐸 𝐶𝑂𝑁𝑄𝑈𝐸𝑆𝑇𝑆𝑖𝑡+𝛽6𝐼𝑀𝑃𝑂𝑅𝑇𝑆𝑖𝑡 + 𝛽𝑡 + 𝜀𝑖𝑡
The results show that in general increments of exports didn’t increase Japanese
productivity. However, there is a perceivable effect at the skill level, although the effect of
increases of high and low skilled exports on productivity is very weak. In other words, the
results in table 10 permit us to reject the possibility that productivity growth was endogenous
to export growth for the case of 1912-1938 Japan.
Another possible source of endogeneity could be present in trade costs in the sense that
their causality relation with exports might be bidirectional. The clearest example is that of
diplomats, since increases in the number of Japanese diplomats in a country might increase
Japanese exports to that country, but the fact that Japan is trading with the country might also
attract a larger number of Japanese consuls there. This possibility is real, as is demonstrated in
table 11, in which the value of Japanese exports affected in a positive and significant way the
number of Japanese Diplomats abroad, the number of foreign diplomats in Japan and total
Diplomatic representation (the sum of diplomats abroad and foreign diplomats in Japan).
Furthermore this phenomenon could also appear in the rest of trade costs or even in productivity
measures.
Table 10: Effect of Changes in Exports on the Number of Japanese Diplomats Abroad
(1) (2) (3) VARIABLES Japan Diplomats abroad Diplomats in Japan TOTAL DIPLOMATS
logexports 0.271*** 0.232*** 0.232***
(0.0433) (0.0221) (0.0395)
Constant -2.731*** -1.622*** -1.537**
(0.669) (0.362) (0.609)
Observations 327 165 327
R-squared 0.170 0.541 0.165
For that reason, it is important to ascertain whether the dependent variable (exports) is
not affecting the independent variables, because it would lead to distorted and biased effects of
trade costs on export growth. In order to do this, we take lags of every independent variable so
as to avoid them being affected by changes in exports, as current exports can’t affect past trade
costs.
76
Table 11: Japanese Exports Determinants (Lagged Independent Variables)
(1)
VARIABLES EXP
GDP 0.977***
(0.300)
RELPRODUCTIVITY 0.544*
(0.292)
RELWAGES 0.125
(0.140)
GDPCAPABSDIFF 1.035**
(0.408)
POPDENSITYDIFF 0.601***
(0.125)
FREIGHTS -0.219
(0.177)
TARIFFS -0.338**
(0.140)
EXCHCONTROL -1.757
(1.097)
EXCHRATE -0.457***
(0.109)
COLONY 2.638***
(0.706)
FUTURE CONQUESTS 0.832***
(0.305)
DIPLOMATS 0.375***
(0.109)
IMPORTS 0.156*
(0.0861)
Constant -4.937
(3.044)
Observations 673
R-squared 0.741
Table 12 shows export determinants lagged by one period, and we can see that the main
results still hold when we account for the effect of trade determinants on exports one period
ahead. By doing this, we avoid bias conclusions generated by the hypothetical effect of export
growth on trade determinants, and we also find that the employed trade determinants were
persistent since changes in productivity and trade costs affected Japanese exports in the future
period.
77
REFERENCES
Allen R.G,D. and Edward Ely (1953): International Trade Statistics. Edt. By Wiley and Sons,
Inc. New York.
Almanach de Gotha. Annuaire généalogique, diplomatique etstatistique. Gotha: Perthes.
Various years.
Bethell, L. -Ed-(1994). The Cambridge History of Latin America.1930 to the Present.Volume
VI, Part 1.Cambridge University Press.
Blattman, C.; Clemens, M. A. Williamson, J.G (2003): «Who Protected and Why? Tariffs the
World around 1870-1938». Harvard Institute of Economic Research Working Papers 2010.
Board of Trade (1905). The comparative incidence of foreign and colonial import tariffs on the
principal classes of manufactures exported from the United Kingdom. British Parliamentary
Papers, 84, 280-320.
Bolt, J. and J. L. van Zanden (2014). The Maddison Project: collaborative research on historical
national accounts. The Economic History Review, 67 (3): 627–651, working paper.
Eichengreen, B. (1996). Golden Fetters: The Gold Standard and the Great Depression, 1919-
1939. Oxford University Press
Ellinger, B., &Ellinger, H. (1930). Japanese competition in the cotton trade.journal of the Royal
Statistical Society, 185-232.
Federico, G. (2010): Feeding the World. An Economic History of Agriculture 1800-2000.
Princeton University Press. Priceton.
Federico, G. and Tena-Junguito, A. (2018): Federico-Tena World Trade Historical Database:
http://www.uc3m.es/tradehist_db
Federico, G. and Tena-Junguito A. (2019): World trade, 1800-1938: a new synthesis. Revista
de Historia Económica-Journal of Iberian and Latin America Economic History, Vol 37, n.1
(forthcomming).
France. Annuaire Statistique de la France, Ministère de l'agriculture et du commerce,
Service de la statistique générale de France by Institutnational de la statistique et des études
économiques (France), Paris.
Government of Taiwan (Taiwan Soutokuhu), 1936. Taiwan Boueki 40 nen hyou (Returns of
the trade of Taiwan for fourty years).
Government of Taiwan (Taiwan Soutokuhu), 1940. Taiwan Gaikoku Boueki Gaikyou (Foreign
Trade of Taiwan).
Huberman, M., Meissner, C. M, and Oosterlinck, K. (2017). Technology and geography in the
second industrial revolution: new evidence from the margins of trade. The Journal of Economic
History, 77(1), 39-89.
78
Huberman, M., &Minns, C. (2007). The times they are not changing’: Days and hours of work
in Old and New Worlds, 1870–2000. Explorations in Economic History, 44(4), 538-567.
Isserlis, L. (1938). Tramp shipping cargoes, and freights. Journal of the Royal Statistical
Society, 101(1), 53-146.
Istat, Italia, & Direzione generale della statistica. (various years). Annuario statistic italiano.
Tip. Elzeviriana.
Kang, C. (1981). Demographic Development in Manchuria, 1924-1941. 近代史研究所集刊,
(10), 351-375.
Kazushi Ohkawa, Miyohei Shinohara, and MatajiUmemura, Long-Term Economic Statistics
(LTES), Toyo Keizai Shinposha, Inc., (1967-1989).
Keller, W., Li, B., & Shiue, C. H. (2011). China’s foreign trade: Perspectives from the past 150
years. The World Economy, 34(6), 853-892.
League of Nations. Economic and Financial Section (1927-1945). Statistical year-book of the
League of Nations.Geneva, [Switzerland: League of Nations, Economic and Financial
Section.].
Maddison, A. (1995). Monitoring the world economy, 1820-1992 (p. 238). Paris: Development
Centre of the Organisation for Economic Co-operation and Development.
Maddison, A. (2007). The world economy volume 1: A millennial perspective volume 2:
Historical statistics. Academic Foundation.
Meisel, A. (1954 - [et al.] (1990). El Banco de la República: antecedentes, evolución y
estructura. Bogotá: Banco de la República. Departamento editorial
Meissner, C. M., &Tang, J. P. (2018): Upstart Industrialization and Exports, Japan 1880-1910).
The Journal of Economic History, Volume 78, Issue 4. December 2018, pp. 1068-1102.
Ministry of Finance.Annual Return of the Foreign Trade of the Empire of Japan. Tokyo:
Yushodo (1912, 1915, 1925, 1929, 1932, 1938).
Mitchell, B. R. (2003). International historical statistics: Africa, Asia & Oceania, 1750–2000.
Houndmills, Basingstokes. Hampshire, New York: Palgrave Macmillan.
Mitchell, B. (2003). International Historical Statistics: the Americas 1750-1993. M.
Mitchell, B. (Ed.). (1998). International historical statistics: Europe 1750-1993. Springer.
Moneta, C. (1959). The estimation of transportation costs in international trade. Journal of
Political Economy, 67(1), 41-58.
Nun, N. and Trefler, D. (2010). The Structure of Tariffs and Long-Term Growth. American
Economic Journal: Macroeconomics Vol. 2 No. 4: 158–194.
Reinhart, C & Rogoff, K. (2002). The Modern History of Exchange Rate Arrangements: A
Reinterpretation. The Quarterly Journal of Economics Vol. CXIX February 2004 Issue 1.
Robertson, A. J. (1990). Lancashire and the Rise of Japan, 1910–1937. Business History, 32(4),
87-105
79
Royal Statistical Office. Statistical Abstract for the Several Colonial and Other Possessions of
the United Kingdom (Statistical Abstract for the British Empire).London: 1880f.
Sanderson, A. Edward. (1926). Ocean freight rates in United States foreign trade. Washington:
Government Printing Office.
Sanderson, A. Edward. (1940). Wartime control of ocean freight rates and foreign trade: a world
survey. Washington: U.S. Govt. print. off.
Saxonhouse, Gary, and Gavin Wright (1984a). "New Evidence on the Stubborn English Mule
and the Cotton Industry, 1878-1920," Economic History Review 37: 507-519.
Saxonhouse, Gary, and Gavin Wright (1984b). "Rings and Mules around the World: A
Comparative Study in Technological Choice," in Saxonhouse and Wright (eds.), Technique,
Spirit and Form in the Making of the Modern Economies: Essays in Honor of William N.
Parker. Research in Economic History, Supplement 3: 271-300.
Shimizu, H. (1988). Dutch-Japanese competition in the shipping trade on the Java-Japan route
in the inter-war period. Japanese Journal of Southeast Asian Studies, 26(1), 3-23.
Tena-Junguito, A. (2010). Bairoch revisited: tariff structure and growth in the late nineteenth
century. European Review of Economic History, 14(1), 111-143.
Williamson, J. G. (1995). The evolution of global labor markets since 1830: background
evidence and hypotheses. Explorations in Economic History, 32(2), 141-196.
Woltcott, S. (1994): The Perils of Lifetime Employment Systems: Productivity Advance in the
Indian and Japanese Textile Industries, 1920-1938. The Journal of Economic History, Vol. 54,
No. 2, pp 307-324.
Yasuba, Y. (1978). Freight rates and productivity in Ocean transportation for Japan, 1875-
1943.Explorations in Economic History, 15(1), 11.
European Historical Economics Society
EHES Working Paper Series Recent EHES Working Papers 2019
EHES 152 Building Workers in Madrid (1737-1805). New Wage Series and Working Lives Mario García-Zúñiga, Ernesto López-Losa
EHES 151 Full steam ahead: Insider knowledge, stock trading and the nationalization of the railways in Prussia around 1879 Michael Buchner, Tobias A. Jopp
EHES 150 Fading Legacies: Human Capital in the Aftermath of the Partitions of Poland Andreas Backhaus
EHES 149 Quantification and Revolution: An Investigation of German Capital Flight after the First World War Christophe Farquet
EHES 148 Without coal in the age of steam and dams in the age of electricity: an explanation for the failure of Portugal to industrialize before the Second World War Sofia Teives Henriques and Paul Sharp
EHES 147 Money and modernization in early modern England Nuno Palma
EHES 146 Class, education and social mobility: Madrid, 1880-1905 Francisco J. Beltrán Tapia and Santiago de Miguel Salanova EHES 145 Is there a Latin American agricultural growth pattern? Factor endowments and productivity in the second half of the twentieth century Miguel Martín-Retortillo, Vicente Pinilla, Jackeline Velazco, Henry Willebald 2018 EHES 144 Peer Pressure: The Puzzle of Tax Compliance in the Early Nineteenth-Century
Russia Elena Korchmina
All papers may be downloaded free of charge from: www.ehes.org The European Historical Economics Society is concerned with advancing education in European economic history through study of European economies and economic history. The society is registered with the Charity Commissioners of England and Wales number: 1052680